The Pitfalls of "Security by Obscurity" And What They Mean for Transparent AI

Olivia Mundahl; Peter Hall; Sunoo Park

REVIEW 3 major objections 5 minor 217 references

The Pitfalls of "Security by Obscurity" And What They Mean for Transparent AI

T0 review · 3 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The security community's rejection of 'security by obscurity' offers AI transparency a ready-made playbook.

desk verdict A competent, honest synthesis of security's transparency norms for AI, with a real but acknowledged disanalogy around metric gaming; deserves peer review. read the letter →

arxiv 2501.18669 v1 pith:I2V6MG2X submitted 2025-01-30 cs.CR cs.AIcs.CY

Peter Hall , Olivia Mundahl , Sunoo Park This is my paper · ORCID

classification cs.CRcs.AIcs.CY

keywords securitybyobscuritytransparencyAIthreatmodelingcoordinatedvulnerabilitydisclosurepublictrustmechanismsanonymizationdifferentialprivacy

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks whether the AI community can learn from the security community's decades-long experience with 'security by obscurity,' the idea that hiding how a system works protects it. It argues that security's hard-won consensus—transparency improves safety, especially when paired with community scrutiny and institutionalized trust—translates into three lessons for AI: state what a system cannot guarantee as clearly as what it can, invite and protect outside researchers who probe for flaws, and build public trust mechanisms for systems people cannot meaningfully opt out of. The paper grounds these lessons in a case study of anonymization, where public disclosure of successful de-anonymization attacks changed industry practice and helped motivate differential privacy. It closes by identifying what is genuinely new in AI—training data enmeshed with model behavior, metric gaming, and model brittleness—so the security playbook needs adaptation rather than simple copying.

What carries the argument

The load-bearing instrument is the analogy between security-critical systems and AI systems, carried by three named mechanisms. The first is threat modeling: the security practice of writing down what a system can and cannot guarantee, so that failures and out-of-scope risks are as explicit as the promised protections. The second is community scrutiny and graded disclosure: bug bounties, coordinated vulnerability disclosure, and attack papers that invite outside eyes and protect researchers who report flaws. The third is public trust mechanisms: regulation, certification, audits, and similar institutional arrangements that substitute for individual understanding when a technology is ubiquitous. The anonymization story—where transparent reports of attacks repeatedly showed that anonymized data could be re-identified, eventually changing industry behavior and helping motivate differential privacy—serves as the paper's concrete demonstration that these mechanisms produce real shifts in practice.

What would settle it

A controlled comparison would settle it: take two matched AI systems with the same capabilities, give one a published threat model, an open disclosure process, and public audits, keep the other closed, and track serious flaws found, fix speed, and exploited harms over a fixed period. If the closed system performs as well or better on these measures, the security-by-transparency parallel fails for that class of systems.

Watch

Extended reading notes

Core claim

The paper's central claim is that transparency debates in AI are replaying the security community's older debate about 'security by obscurity,' and the security community's eventual consensus gives AI a useful template. That consensus, called here 'security by transparency,' holds that a system should be designed to remain safe even when its design is fully known, and that openness about assumptions, vulnerabilities, and guarantees makes systems safer over time. The three transferable mechanisms are threat modeling that specifies both positive and negative guarantees, community input through vulnerability reporting and graded disclosure, and public trust mechanisms that let lay users rely on systems they cannot personally audit. The anonymization case study is the paper's worked example: repeated transparent demonstrations that anonymization fails changed norms and led to differential privacy. The paper also argues that AI differs in ways that matter for the analogy—training data cannot be cleanly separated from system design, published metrics can be gamed, and models are brittle—so the template must be adapted to these conditions.

Load-bearing premise

The transfer works only if AI systems and security-critical systems are similar enough that publishing negative guarantees, inviting outside scrutiny, and building public trust mechanisms yield the same safety benefits in AI that they produced in security.

Editorial extensions

If this is right

Publishing a model's threat model, including explicit statements of what it does not guarantee, should become a standard complement to benchmarks and capability claims.
AI developers who build and respect coordinated disclosure processes, with legal protection for researchers, should see faster discovery of serious flaws, just as security did.
Regulators can justify transparency requirements even when users do not demand them, because ubiquitous AI creates the same moral-hazard conditions that led security to institutionalize public trust mechanisms.
Because AI has challenges security did not face, the playbook must be adapted—for example, disclosure processes must decide who counts as the affected party when training data is entangled with model behavior.
The anonymization history shows that sustained transparent reporting of failures can shift community practice, so AI transparency advocates should expect such change to take time.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Beyond the paper, the threat-modeling lesson implies that AI audits should carry a standard 'what was not tested' section, making audits comparable across models and vendors.
Beyond the paper, the metric-gaming problem may be the deepest limit of the analogy: publishing an objective function can make it a target in a way that publishing a cryptographic design does not, so effective AI transparency may need to publish evaluation protocols rather than the metrics themselves.
Beyond the paper, if the security timeline repeats, the AI field should expect a transitional period of hostile reactions to outside researchers followed by institutionalized disclosure norms, so early investment in safe-harbor structures could compress that process.
Beyond the paper, a testable extension is to track whether AI vulnerability reports using coordinated disclosure lead to faster fixes and fewer exploited incidents than unsolicited public reports; the paper's analogy predicts the former.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

This is a systematic synthesis, not a breakthrough. It reorganizes known security principles—Kerckhoffs's principle, coordinated disclosure, public trust mechanisms—into three themes and maps them onto AI transparency. The contribution is the framing and the cross-domain transfer, and by that standard it works: the paper is well structured, well cited, and unusually honest about where the analogy frays.

What it does well: the three parallels are clear and concrete. Section 3.1's emphasis on negative guarantees (threat models as specifications of what is not covered) is a useful lens for AI, where much current work focuses on positive guarantees. The anonymization case study in Section 5 shows how transparency norms emerged from harm-driven disclosure and gives the argument an empirical anchor. The paper also carefully distinguishes its scope—transparency, not security—and flags that the security community's own norms are imperfect and ongoing. That kind of self-limitation is rare and worth respecting.

Soft spots: the three themes are selected without a stated method; they are a reasonable editorial choice, but the paper would be stronger if it explained why these three and not others. The 'many eyes' assumption is treated as largely valid. It does cite Schneier's caveat (open source only helps if people actually read the code) and Sharma's incentive critique, but the core parallel still leans on Linus's Law more than the evidence fully supports. Finally, the Goodhart disanalogy in Section 6.4 is real: disclosing optimization metrics can invite gaming, a failure mode with no security analogue. The paper acknowledges this, calls it a difference in kind, and then asserts that the transparency approach would still suggest benefits—without a mechanism for why. That is a genuine open problem, not a hidden circularity; the stress-test's charge of circularity overstates. But the central transfer claim would be more compelling with at least a sketch of how metric disclosure could be managed (e.g., graded disclosure, or transparency at a level that does not enable gaming).

The paper is for AI transparency researchers, security researchers curious about the analogy, and policy people seeking a structured comparison. It deserves a serious referee; it is a competent and useful synthesis that could be improved with a clearer selection rationale and more engagement with the metrics tension.

Recommendation: send to peer review, with revisions. I would be inclined to accept after those refinements.

Referee Report

3 major / 5 minor

Summary. The paper is a position/survey essay arguing that the computer security community's decades-long transition away from "security by obscurity" and toward "security by transparency" provides transferable lessons for contemporary AI transparency debates. It identifies three themes from security practice: (1) explicit threat modeling that states both positive and negative guarantees, (2) "many eyes" community scrutiny through vulnerability disclosure and bug bounties, and (3) public trust mechanisms for ubiquitous technologies whose users cannot opt out. It illustrates these themes with a case study on anonymization, then examines four ways in which AI systems differ from traditional security-critical systems (training data entanglement, disclosure scope, model brittleness/privacy trade-offs, and metric gaming through Goodhart's Law), and finally discusses common arguments against transparency such as trade secrets and misuse. The paper makes no quantitative claims and its argument rests on historical narrative, community practice, and cited literature.

Significance. If the transfer argument holds, the AI transparency community gains a concrete, field-tested template: model what a system cannot guarantee, institutionalize community scrutiny through graded disclosure, and build public trust mechanisms. The paper's strengths are its careful historical grounding, its extensive and relevant bibliography, and its explicit acknowledgment of disanalogies between security and AI. The anonymization case study is a useful illustration of how transparency norms develop in a subfield. The paper is best read as an agenda-setting contribution rather than a technical result; its value lies in framing research and policy questions, and it is honest about several open problems. The main risk is that the central analogy is asserted more strongly than the evidence warrants, particularly around the "many eyes" mechanism and the Goodhart disanalogy.

major comments (3)

[Section 6.4 (Optimizing for Metrics)] The paper identifies Goodhart's Law as a disanalogy with no good analogue in security research, stating that "releasing the full details of system functionality can inherently make the system less useful for its intended purpose." Yet the same paragraph then asserts that "the security-by-transparency approach would suggest that transparency of metrics yields important benefits," without providing a mechanism by which community scrutiny could outweigh metric gaming. This is a load-bearing logical gap: if disclosing metrics creates a qualitatively new failure mode, the many-eyes benefit cannot be assumed to survive. Please either supply an argument or evidence that the benefits of transparency outweigh gaming in the AI context, or explicitly weaken the claim to state that this particular parallel is not established and that the transfer argument applies with appropriate caveats. The current text leaves the reader with an unresolved contradiction at the heart of the second theme.
[Section 3.2 (Many Eyes and Disclosure)] The paper adopts Linus's Law ("Given enough eyeballs, all bugs are shallow") as the underpinning of the many-eyes theme without critically assessing its empirical status in security. It cites Schneier's caveat that open source only increases security if people actually study the code and fix bugs promptly, and it cites Sharma's concern about adverse incentives in decentralized open source, but these caveats are not integrated into the argument. Since the entire second theme depends on the many-eyes mechanism, the paper should either present supporting evidence from security practice (e.g., data from bug bounty programs or cryptographic standardization efforts) or weaken the theme to "community scrutiny can help, subject to known conditions such as active review and aligned incentives." The transfer to AI is only as strong as this premise, so the current treatment is insufficiently qualified.
[Section 1.3 and Section 3] The paper describes its three themes as "key" and offers "a novel and systematic exposition of the essential principles," but it does not state any criteria for selecting these three themes from the large body of transparency literature in security. The lack of an explicit inclusion method makes it difficult to assess whether the three themes are representative or selectively chosen. This concern is amplified by Section 6, which introduces four disanalogies (training data entanglement, disclosure scope, brittleness, metric gaming) without mapping them back to the three themes and without explaining how each theme's validity is affected. Please add a short statement of how the themes were selected and an explicit discussion of which Section 6 challenges threaten which theme.

minor comments (5)

[Section 7.3] There is a typo in the sentence "reach fuller understandings of of new techniques," which should read "of new techniques."
[References] The reference to "Fanstastic Copyrighted Beasts" in the GenLaw paper title should be corrected to "Fantastic Copyrighted Beasts."
[Section 5] The anonymization case study is informative, but its connection to the three themes is mostly implicit; explicitly labeling which paragraphs illustrate threat modeling, many eyes, and public trust would strengthen the didactic value of the case study.
[Section 8.1] The discussion of comparisons to biosecurity, physical sciences, biomedical sciences, and civil engineering is quite brief and does not fully explain how each comparison informs the central transfer argument; consider either expanding these paragraphs with concrete parallels or trimming them to avoid distracting from the main narrative.
[Section 1.4] The distinction between "transparency" and "security" is useful, but the paper sometimes conflates the two, e.g., in Section 3.3 where "transparency is necessary as a public trust mechanism for secure systems and robust AI alike"; a brief clarifying sentence would help avoid ambiguity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's analogical synthesis rests on external security history and explicitly acknowledges AI-specific disanalogies.

full rationale

This is an argumentative and historical position paper with no quantitative derivations, fitted parameters, or formal predictions. Its central claim—that the security community's transparency norms offer transferable lessons for AI—rests on documented external history (Kerckhoffs, Shannon, Raymond, NIST standardization, vulnerability disclosure practices, the anonymization case study) rather than on the paper's own outputs. The three key parallels in Section 3 are each supported by citations to independent, publicly documented practices; they are not defined in terms of the conclusions they are used to support. The paper's self-citations ([94] and [138]) are background references for existing concerns and legal risks in research; neither is load-bearing for the central thesis, and neither is invoked as an external authority or uniqueness theorem. Section 6 explicitly identifies AI-specific disanalogies—training-data entanglement, model brittleness, and Goodhartian metric gaming—and even states that the last 'does not have a good analogue in security research.' This candor undercuts any suggestion that the paper is forcing its conclusion by construction; the acknowledged limitations are arguments about the strength of the analogy, not circular derivations. The apparent circularity concern at Section 6.4 is a challenge to the plausibility of an inference from the security-by-transparency paradigm, not a reduction of the conclusion to its premises. No step in the paper equates an input with an output by definition, fits a parameter and then relabels it as a prediction, or imports a result solely through a self-citation chain. The analysis is self-contained as an interpretive synthesis of external evidence, so the circularity score is 0.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The paper introduces no free parameters and no invented entities. The argument depends on four domain assumptions about the reliability and transferability of security transparency norms, listed above. These assumptions are plausible but not formally proven.

assumptions (4)

domain assumption The security community's long-standing consensus against security by obscurity is a transferable source of wisdom for AI transparency debates.
The paper's central thesis assumes that security's historical experience is relevant and reliable for AI; Section 1.1 and Section 3 state this framing.
domain assumption Transparency generally improves security outcomes, with obscurity as at most a minor supplement.
Treated as established by Kerckhoffs' Principle and community consensus; Section 2.1. This claim is background for all three themes.
domain assumption AI systems and security-critical systems share enough structural features for the three parallels to hold.
Section 3 draws the parallels; Section 6 acknowledges but only partially explores differences such as training data entanglement and metric gaming.
domain assumption Community scrutiny, as expressed in Linus's Law, reliably finds and fixes flaws.
Section 3.2 invokes 'many eyes' as a key mechanism; the paper cites Schneier's caveats but does not systematically test or qualify the claim.

how reviews work

0 comments

Cite this review

Pith. "Pith review of The Pitfalls of "Security by Obscurity" And What They Mean for Transparent AI." pith.science (2026). https://pith.science/paper/I2V6MG2X

@misc{pith2026250118669,
  author       = {Pith},
  title        = {Pith review of: The Pitfalls of "Security by Obscurity" And What They Mean for Transparent AI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I2V6MG2X}},
  note         = {Machine review of arXiv:2501.18669}
}

read the original abstract

Calls for transparency in AI systems are growing in number and urgency from diverse stakeholders ranging from regulators to researchers to users (with a comparative absence of companies developing AI). Notions of transparency for AI abound, each addressing distinct interests and concerns. In computer security, transparency is likewise regarded as a key concept. The security community has for decades pushed back against so-called security by obscurity -- the idea that hiding how a system works protects it from attack -- against significant pressure from industry and other stakeholders. Over the decades, in a community process that is imperfect and ongoing, security researchers and practitioners have gradually built up some norms and practices around how to balance transparency interests with possible negative side effects. This paper asks: What insights can the AI community take from the security community's experience with transparency? We identify three key themes in the security community's perspective on the benefits of transparency and their approach to balancing transparency against countervailing interests. For each, we investigate parallels and insights relevant to transparency in AI. We then provide a case study discussion on how transparency has shaped the research subfield of anonymization. Finally, shifting our focus from similarities to differences, we highlight key transparency issues where modern AI systems present challenges different from other kinds of security-critical systems, raising interesting open questions for the security and AI communities alike.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

217 extracted references · 30 canonical work pages

[1]

Artiﬁcial Intelligence and Jobs — Evidence from Online Vacancies

Daron Acemoglu, David Autor, Jonathon Hazell, and Pascual Restrepo. “Artiﬁcial Intelligence and Jobs — Evidence from Online Vacancies”. In: Journal of Labor Economics 40 (1) (2022), pp. 293–340. doi: 10.1086/718327. url: https:// economics.mit.edu/sites/default/ files/publications/AI%20and%20Jobs %20-%20Evidence%20from%20Online %20Vacancies.pdf

doi:10.1086/718327 2022
[2]

RSA Factor- ing Challenge Announcement

RSA Challenge Administrator. RSA Factor- ing Challenge Announcement . permalink: https://perma.cc/Z3W7-FCDU. 1991. url: https://groups.google.com/u/1/g/sci. crypt/c/AA7M9qWWx3w/m/EkrsR69CDqIJ? pli=1

1991
[3]

Infor- mation Sharing

America’s Cyber Defense Agency. Infor- mation Sharing . permalink: https:// perma.cc/3GP9-MD94. url: https://www. cisa.gov/topics/cyber-threats-and- advisories/information-sharing
[4]

Perspectives on Issues in AI Governance

Google AI. Perspectives on Issues in AI Governance . Tech. rep. Google,
[5]

Ethics Guidelines for Trustworthy AI

High-Level Expert Group on AI. Ethics Guidelines for Trustworthy AI . European Commission. 2019

2019
[6]

Piloting a survey- based assessment of transparency and trustworthiness with three medical AI tools

Jana Fehr et al. “Piloting a survey- based assessment of transparency and trustworthiness with three medical AI tools”. In: Healthcare (Basel) (2022). doi: 10.3390/healthcare10101923. url: https://www.ncbi.nlm.nih.gov/pmc/ articles/PMC9601535/

doi:10.3390/healthcare10101923 2022
[7]

Apple pays hackers six ﬁgures to ﬁnd bugs in its software. Then it sits on their ﬁndings

Reed Albergotti. “Apple pays hackers six ﬁgures to ﬁnd bugs in its software. Then it sits on their ﬁndings.” In: Washing- ton Post (2021). permalink: https:// perma.cc/5LTD-4YBW. url: https://www. washingtonpost.com/technology/2021/ 09/09/apple-bug-bounty/

2021
[8]

The Falcon Series of 14 Open Language Models

Ebtesam Almazrouei, Hamza Alobeidli, Abdulaziz Alshamsi, Alessandro Cappelli, Ruxandra Cojocaru, M´ erouane Debbah, Etienne Goﬃnet, Daniel Hesslow, Julien Launay, Quentin Malartic, Daniele Mazzotta, Badreddine Noune, Baptiste Pannier, and Guilherme Penedo. “The Falcon Series of 14 Open Language Models”. In: Computation and Language (2023). url: https://ar...

arXiv 2023

Show all 217 references

[9]

Security Engineering

Ross Anderson. Security Engineering. 3rd ed. Wiley, 2021

2021
[10]

Varieties of transparency: exploring agency within AI systems

Gloria Andrada, Robert W. Clowes, and Paul R. Smart. “Varieties of transparency: exploring agency within AI systems”. In: AI and SOCIETY 38 (2022), pp. 1321–

2022
[11]

How to De-Identify your Data

Olivia Angiuli, Joe Blitzstein, and Jim Waldo. “How to De-Identify your Data”. In: Communications of the ACM (2015). permalink: https://perma.cc/BY4P-828Y. url: https://cacm.acm.org/practice/ how-to-de-identify-your-data/

2015
[12]

Machine Bias

Julia Angwin, Jeﬀ Larson, Surya Mattu, and Lauren Kirchner. Machine Bias . ProP- ublica. permalink: https://perma. cc/V5XH-GMQ8. 2016. url: https:// www.propublica.org/article/machine- bias-risk-assessments-in-criminal- sentencing

2016
[13]

Argentieri

Nicole M. Argentieri. Principal Deputy Assis- tant Attorney General Nicole M. Argentieri Delivers Remarks at the Computer Crime and Intellectual Property Section’s Symposium on Artiﬁcial Intelligence in the Justice Dept. U.S. Dept. of Justice. permalink: https:// perma.cc/3PCD...

2024
[14]

Artiﬁcial Intelligence Risk Management Framework: Generative Artiﬁcial In- telligence Proﬁle . NIST. 2024. doi: 10.6028/NIST.AI.600-1. url: https:// doi.org/10.6028/NIST.AI.600-1

2024 doi
[15]

Theoretically Grounded Loss Functions and Algorithms for Adversarial Robustness

Pranjal Awasthi, Anqi Mao, Mehryar Mohri, and Yutao Zhong. “Theoretically Grounded Loss Functions and Algorithms for Adversarial Robustness”. In: Proceed- ings of The 26th International Conference on Artiﬁcial Intelligence and Statistics . Vol. 206. Proceedings of Machine Lear...

2023
[16]

A Face Is Exposed for AOL Searcher No

Michael Barbaro and Tom Zeller Jr. A Face Is Exposed for AOL Searcher No. 4417749 . permalink: https:// archive.nytimes.com/www.nytimes.com/ learning/teachers/featured_articles/ 20060810thursday.html. 2006

2006
[17]

A Feeling of Unease about Privacy Law

Ann Bartow. “A Feeling of Unease about Privacy Law”. In: University of Pennsylvania Law Review Online (2006). url: https:// scholarship.law.upenn.edu/penn_law_ review_online/vol155/iss1/6

2006
[18]

The Re- search Value of Publishing Attacks

David Basin and Srdjan Capkun. “The Re- search Value of Publishing Attacks”. In: Com- munications of the ACM 55 (11 2012), pp. 22–

2012
[19]

The Artiﬁcial Intelligence Black Box and the Failure of Intent and Cau- sation

Yavar Bathaee. “The Artiﬁcial Intelligence Black Box and the Failure of Intent and Cau- sation”. In: Harvard Journal of Law and Tech- nology 31 (2 2018)

2018
[20]

Bellovin and Randy Bush

Steven M. Bellovin and Randy Bush. Security Through Obscurity Considered Dangerous. The Internet Society. 2002. url: https://www.cs.columbia.edu/~smb/ papers/draft-ymbk-obscurity-00.txt

2002
[21]

On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?

Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?” In: Pro- ceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (F AccT ’21). ACM, 2021, pp. 61...

2021
[22]

Managing extreme AI risks amid rapid progress

Yoshua Bengio, Geoﬀrey Hinton, Andrew Yao, Dawn Song, Pieter Abbeel, Trevor Darrell, Yuval Noah Harari, Ya-Qin Zhang, Lan Xue, Shai Shalev-Shwartz, Gillian Hadﬁeld, Jeﬀ Clune, Tegan Maharaj, Frank Hutter, Atlim G¨ unes, Baydin, Sheila McIlraith, Qiqi Gao, Ashwin Acharya, David...

2024
[23]

Joseph R. Biden. Executive Order on Preventing Access to Americans’ Bulk Sen- sitive Personal Data and United States Government-Related Data by Coun- tries of Concern . 2024. url: https:// web.archive.org/web/20250118021320/ 15 https://www.whitehouse.gov/briefing- room/preside...

2024
[24]

doi: 10.1145/2366316.2366324
[25]

International Conference on Machine Learn- ing

Pythia: A Suite for Analyzing Large Lan- guage Models Across Training and Scaling . International Conference on Machine Learn- ing. 2023. url: https://proceedings.mlr. press/v202/biderman23a.html

2023
[26]

Semantically secure order-revealing encryption: Multi-input functional encryption without obfuscation

Dan Boneh, Kevin Lewi, Mariana Raykova, Amit Sahai, Mark Zhandry, and Joe Zim- merman. “Semantically secure order-revealing encryption: Multi-input functional encryption without obfuscation”. In: EUROCRYPT. EU- ROCRYPT, 2015, pp. 563–594

2015
[27]

Joseph R. Biden. Executive Order on the Safe, Secure, and Trustworthy Develop- ment and Use of Artiﬁcial Intelligence . EO14110. permalink: https://perma. cc/LPA7-QWTQ. 2023. url: https://www. federalregister.gov/documents/2023/ 11/01/2023-24283/safe-secure-and- trustworthy-de...

2023
[28]

Strategic Implications of Openness in AI Development

Nick Bostrom. “Strategic Implications of Openness in AI Development”. In: Global Pol- icy (2017). doi: 10.1111/1758-5899.12403. url: https://doi.org/10.1111/1758- 5899.12403

2017
[29]

Strategic Implications of Openness in AI Development

Nick Bostrom. “Strategic Implications of Openness in AI Development”. In: Artiﬁcial Intelligence Safety and Security (2018). url: https://nickbostrom.com/papers/ openness.pdf

2018
[30]

The Science of Guessing: Analyzing an anonymized corpus of 70 million passwords

Joseph Bonneau. “The Science of Guessing: Analyzing an anonymized corpus of 70 million passwords”. In: 2012 IEEE Symposium on Security and Privacy . IEEE Symposium on Security and Privacy, 2012. url: https:// ieeexplore.ieee.org/stamp/stamp.jsp? tp=&arnumber=6234435&tag=1

2012
[31]

Miles Brundage, Shahar Avin, Jack Clark, Helen Toner, Peter Eckersley, Ben Garﬁnkel, Allan Dafoe, Paul Scharre, Thomas Zeitzoﬀ, Bobby Filar, Hyrum Anderson, Heather Roﬀ, Gregory C. Allen, Jacob Steinhardt, Carrick Flynn, Se´ an ´O h ´Eigeartaigh, Simon Beard, Haydn Belﬁeld, Se...

2018
[32]

On Eval- uating Adversarial Robustness

Nicholas Carlini, Anish Athalye, Nicolas Pa- pernot, Wieland Brendel, Jonas Rauber, Dim- itris Tsipras, Ian Goodfellow, Aleksander Madry, and Alexey Kurakin. “On Eval- uating Adversarial Robustness”. In: arXiv (2019). doi: 10.48550/arXiv.1902.06705. url: https://arxiv.org/abs/...

2019 doi arXiv
[33]

Artiﬁcial Bias: The Ethical Concerns of AI-Driven Dispute Resolution in Family Matters

Wensdai Brooks. “Artiﬁcial Bias: The Ethical Concerns of AI-Driven Dispute Resolution in Family Matters”. In: Journal of Dispute Reso- lution (2 2022). url: https://scholarship. law.missouri.edu/jdr/vol2022/iss2/9

2022
[34]

Stealing Part of a Production Language Model

Nicholas Carlini, Daniel Paleka, Krishna- murthy Dj Dvijotham, Thomas Steinke, Jonathan Hayase, A. Feder Cooper, Katherine Lee, Matthew Jagielski, Milad Nasr, Arthur Conmy, Itay Yona, Eric Wallace, David Rol- nick, and Florian Tram` er. “Stealing Part of a Production Language ...

2024
[35]

The Need for Trans- parency in the Age of Predictive Sentencing Algorithms

Alyssa M. Carlson. “The Need for Trans- parency in the Age of Predictive Sentencing Algorithms”. In: Iowa Law Review 103 (2017). url: https://ilr.law.uiowa.edu/ sites/ilr.law.uiowa.edu/files/2022- 10/The%20Need%20for%20Transparency %20in%20the%20Age%20of%20Predictive %20Senten...

2017
[36]

Extracting Training Data from Diﬀusion Models

Nicholas Carlini, Jamie Hayes, Milad Nasr, Matthew Jagielski, Vikash Sehwag, Florian Tram` er, Borja Balle, Daphne Ippolito, and Eric Wallace. “Extracting Training Data from Diﬀusion Models”. In: arXiv (2023). doi: 10.48550/ARXIV.2301.13188. url: https://doi.org/10.48550/ARXIV...

2023 doi arXiv
[37]

An ef- ﬁcient key recovery attack on SIDH

Wouter Castryck and Thomas Decru. “An ef- ﬁcient key recovery attack on SIDH”. In: EU- ROCRYPT 2023. Springer, 2023, pp. 423–447. doi: 10.1007/978-3-031-30589-4_15 . url: https://eprint.iacr.org/2022/975

2023 doi
[38]

Coordinated Flaw Disclosure for AI: Beyond Security Vulnerabilities

Sven Cattell, Avijit Ghosh, and Lucie-Aim´ ee Kaﬀee. “Coordinated Flaw Disclosure for AI: Beyond Security Vulnerabilities”. In: AAAI/ACM Conf. on AI, Ethics, and So- ciety (AIES) (2024). url: https://arxiv. org/pdf/2402.07039

2024 arXiv
[39]

Towards a Data Privacy- Predictive Performance Trade-oﬀ

Tˆ ania Carvalho, Nuno Moniz, Pedro Faria, and Lu ´ ıs Antunes. “Towards a Data Privacy- Predictive Performance Trade-oﬀ”. In: Expert Systems with Applications 223 (2023). doi: 10.1016/j.eswa.2023.119785. url: https://doi.org/10.1016/j.eswa.2023. 119785. 16

2023
[40]

The Eﬀect of Internet Security Breach Announcements on Market Value: Capital Market Reactions for Breached Firms and Internet Security Developers

Huseyin Cavusoglu, Birendra Mishra, and Srinivasan Raghunathan. “The Eﬀect of Internet Security Breach Announcements on Market Value: Capital Market Reactions for Breached Firms and Internet Security Developers”. In: International Journal of Electronic Commerce 9 (2004), pp. 7...

2004
[41]

CERT Guide to Coor- dinated Vulnerability Disclosure

CERT. CERT Guide to Coor- dinated Vulnerability Disclosure . https://certcc.github.io/CERT-Guide-to- CVD/topics/phases/deployment/. perma- link: https://perma.cc/6YE9-FMK4
[42]

Fac- torization of a 512-Bit RSA Modulus

Stefania Cavallar, Bruce Dodson, Arjen K. Lenstra, Walter Lioen, Peter L. Mont- gomery, Brian Murphy, Herman te Riele, Karen Aardal, Jeﬀ Gilchrist, G´ erard Guillerm, Paul Leyland, J¨ oel Marchand, Fran¸ cois Morain, Alec Muﬀett, Chris Putnam, Craig Putnam, and Paul Zimmermann...

doi
[43]

A wave of retractions is shaking physics

Sophia Chen. A wave of retractions is shaking physics. permalink: https://perma.cc/ AXC6-FPZF. MIT Technology Review, 2024. url: https://www.technologyreview. com/2024/05/15/1092535/a-wave-of- retractions-is-shaking-physics/? truid=&utm_source=the_download& utm_medium=email&ut...

2024
[44]

Pseu- dorandom Error-Correcting Codes

Miranda Christ and Sam Gunn. “Pseu- dorandom Error-Correcting Codes”. In: CRYPTO 2024 . 2024. doi: 10.48550/arXiv.2402.09370. url: https://eprint.iacr.org/2024/235. pdf

2024 doi arXiv
[45]

permalink: https://perma

CISA Open Source Software Security Roadmap. permalink: https://perma. cc/N26U-GT2G. 2023. url: https://www. cisa.gov/sites/default/files/2024-02/ CISA-Open-Source-Software-Security- Roadmap-508c.pdf

2023
[46]

Security through obscurity: An approach for protecting Register Transfer Level hard- ware IP

Rajat Subhra Chakraborty and Swarup Bhu- nia. “Security through obscurity: An approach for protecting Register Transfer Level hard- ware IP”. In: IEEE International Workshop on Hardware-Oriented Security and Trust, 2009. doi: 10.1109/HST.2009.5224963

2009
[47]

Attacks on Deidentiﬁcation’s Defenses

Aloni Cohen. “Attacks on Deidentiﬁcation’s Defenses”. In: 31st USENIX Security Sym- posium, USENIX Security 2022, Boston, MA, USA, August 10-12, 2022 . Ed. by Kevin R. B. Butler and Kurt Thomas. USENIX Association, 2022, pp. 1469–1486. url: https://www.usenix.org/conference/ u...

2022
[48]

Ac- countability in an Algorithmic Society: Re- lationality, Responsibility, and Robustness in Machine Learning

A. Feder Cooper, Emanuel Moss, Ben- jamin Laufer, and Helen Nissenbaum. “Ac- countability in an Algorithmic Society: Re- lationality, Responsibility, and Robustness in Machine Learning”. In: ACM Conf. on Fairness, Accountability, and Transparency (2022), pp. 864–876

2022
[49]

Synthetic Data: Methods, Use Cases, and Risks

Emiliano De Cristofaro. “Synthetic Data: Methods, Use Cases, and Risks”. In: IEEE Secur. Priv. 22.3 (2024), pp. 62–

2024
[50]

Stan- dards

American Society of Civil Engineers. Stan- dards. permalink: https://perma.cc/ 23YQ-EJS3. url: https://ascelibrary. org/topic/ascebookseries/standards
[51]

It’s Time to End Information Anarchy

Scott Culp. It’s Time to End Information Anarchy. Microsoft Technet essay. 2001. url: https://web.archive.org/web/ 20040227175033/http:/www.microsoft. com/technet/columns/security/essays/ noarch.asp. 17

2001
[52]

Searchable symmetric encryption: improved deﬁnitions and eﬃcient constructions

Reza Curtmola, Juan A. Garay, Seny Kamara, and Rafail Ostrovsky. “Searchable symmetric encryption: improved deﬁnitions and eﬃcient constructions”. In: ACM Conf. on Computer and Communications Security (CCS), 2006, pp. 79–88

2006
[53]

Building Guardrails in AI Systems with Threat Modeling

Jayati Dev, Nuray Akhuseyinoglu, Golam Kayas, Bahman Rashidi, and Vaibhav Garg. “Building Guardrails in AI Systems with Threat Modeling”. In: Digital Govern- ment: Research and Practice (2024). doi: 10.1145/3674845. url: https://doi.org/ 10.1145/3674845

2024 doi
[54]

Risky Business: Keeping Security a Secret

Whitﬁeld Diﬃe. Risky Business: Keeping Security a Secret . ZDNet. permalink: https://perma.cc/9BFA-QXNX. 2003. url: https://www.zdnet.com/article/risky- business-keeping-security-a-secret- 5000127072/

2003
[55]

Upholding the Pub- lic Trust

Jim H. Crumpacker. “Upholding the Pub- lic Trust”. In: Public Roads 71 (4 2008). url: https://highways.dot.gov/public- roads/janfeb-2008/upholding-public- trust

2008
[56]

Revealing in- formation while preserving privacy

Irit Dinur and Kobbi Nissim. “Revealing in- formation while preserving privacy”. In: ACM SIGACT-SIGMOD-SIGART Symposium on Principles of Database Systems . Proceed- ings of the Twenty-Second ACM SIGACT- SIGMOD-SIGART Symposium on Principles of Database Systems, 2003, pp. 202–210

2003
[57]

The Menlo Report

David Dittrich and Erin Kenneally. The Menlo Report. Tech. rep. U.S. Department of Home- land Security, 2012. url: https://www.dhs. gov/sites/default/files/publications/ CSD-MenloPrinciplesCORE-20120803_1. pdf

2012
[58]

Diﬀerential Privacy

Cynthia Dwork. “Diﬀerential Privacy”. In: International Conf. on Automata, Languages, and Programming (ICALP) (2006), pp. 1–

2006
[59]

Calibrating noise to sensitivity in private data analysis

Cynthia Dwork, F. McSherry, K. Nissim, and A. Smith. “Calibrating noise to sensitivity in private data analysis”. In: Theory of Cryptog- raphy Conference ’06 . Theory of Cryptogra- phy Conference (TCC), 2006, pp. 265–284

2006
[60]

Tor: The second-generation onion router

Roger Dingledine, Nick Mathewson, and Paul Syverson. “Tor: The second-generation onion router”. In: Proceedings of the 13th conference on USENIX Security Symposium . Vol. 13. ACM, 2004. doi: 10.5555/1251375.1251396

2004
[61]

The Algo- rithmic Foundations of Diﬀerential Privacy

Cynthia Dwork and Aaron Roth. “The Algo- rithmic Foundations of Diﬀerential Privacy”. In: Foundations and Trends in Theoretical Computer Science 9 (3-4 2014), pp. 211–407. doi: 10.1561/0400000042

2014 doi
[62]

Hacking the Law: Are Bug Bounties a True Safe Harbor?

Amit Elazari. “Hacking the Law: Are Bug Bounties a True Safe Harbor?” In: Enigma. Santa Clara, CA: USENIX, 2018. url: https://www.usenix.org/node/208178

2018
[63]

Transparency and the Black Box Problem: Why We Do Not Trust AI

Warren J. von Eschenbach. “Transparency and the Black Box Problem: Why We Do Not Trust AI”. In: Philosophy and Technology 34 (2021), pp. 1607–1622. doi: 10.1007/s13347-021-00477-0 . url: https://doi.org/10.1007/s13347-021- 00477-0

2021 doi
[64]

url: https://link.springer.com/ chapter/10.1007/11787006_1

doi
[65]

Red-Teaming for Generative AI: Silver Bullet or Security Theater?

Michael Feﬀer, Anusha Sinha, Wesley H. Deng, Zachary C. Lipton, and Hoda Heidari. “Red-Teaming for Generative AI: Silver Bullet or Security Theater?” In: Computers and So- ciety (2024). url: https://arxiv.org/pdf/ 2401.15897

2024 arXiv
[66]

Privacy- preserving datamining on vertically parti- tioned databases

Cynthia Dwork and Kobbi Nissim. “Privacy- preserving datamining on vertically parti- tioned databases”. In: CRYPTO. CRYPTO, 2004, pp. 528–544

2004
[67]

url: https://doi.org/10.1109/MSEC.2024

doi: 10.1109/MSEC.2024.3371505. url: https://doi.org/10.1109/MSEC.2024. 3371505

2024
[68]

Cryptography Engineering: Design Principles and Practical Applications

Niels Ferguson, Bruce Schneier, and Ta- dayoshi Kohno. Cryptography Engineering: Design Principles and Practical Applications . John Wiley and Sons, 2010

2010
[69]

Open sourcing AI: intellectual property at the service of platform leader- ship

Carlos Mu˜ noz Ferrandis and Marta Duque Lizarralde. “Open sourcing AI: intellectual property at the service of platform leader- ship”. In: JIPITEC (2022). 18

2022
[70]

Ask an Ethicist: Vulnera- bility Disclosure

ACM Ethics. Ask an Ethicist: Vulnera- bility Disclosure . permalink: https:// perma.cc/3XXU-LXPU. url: https:// ethics.acm.org/integrity-project/ ask-an-ethicist/ask-an-ethicist- vulnerability-disclosure/
[71]

Judicial leadership matters (yet again): the asso- ciation between judge and public trust for artiﬁcial intelligence in courts

Anna Fine and Shawn Marsh. “Judicial leadership matters (yet again): the asso- ciation between judge and public trust for artiﬁcial intelligence in courts”. In: Discover Artiﬁcial Intelligence 4 (2024). doi: 10.1007/s44163-024-00142-3 . url: https://doi.org/10.1007/s44163-024- 00142-3

2024 doi
[72]

A trustworthy AI reality-check: the lack of trans- parency of artiﬁcial intelligence products in healthcare

Jana Fehr, Brian Citro, Rohit Malpani, Christoph Lippert, and Vince I. Madai. “A trustworthy AI reality-check: the lack of trans- parency of artiﬁcial intelligence products in healthcare”. In: Frontiers in Digital Health 6 (2024). doi: 10.3389/fdgth.2024.1267290

2024
[73]

Towards Transparency by Design for Artiﬁcial Intelligence

Heike Felzmann, Eduard Fosch-Villaronga, Christoph Lutz, and Aurelia Tam` o-Larrieux. “Towards Transparency by Design for Artiﬁcial Intelligence”. In: Science and En- gineering Ethics 26 (2020), pp. 3333–3361. doi: 10.1007/s11948-020-00276-4 . url: https://doi.org/10.1007/s119...

2020 doi
[74]

Electronic Frontier Foundation. MBTA v. Anderson. url: https://www.eff.org/ cases/mbta-v-anderson
[75]

Electronic Frontier Foundation. U .S. v. Auernheimer. url: https://www.eff.org/ cases/us-v-auernheimer
[76]

Fairness and Bias in Artiﬁ- cial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies

Emilio Ferrara. “Fairness and Bias in Artiﬁ- cial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies”. In: Sci 6 (1 2023). doi: 10.3390/sci6010003. url: https://doi.org/10.3390/sci6010003

2023 doi
[77]

BadLlama: cheaply removing safety ﬁne-tuning from Llama 2-Chat 13B

Pranav Gade, Simon Lermen, Charlie Rogers- Smith, and Jeﬀrey Ladish. “BadLlama: cheaply removing safety ﬁne-tuning from Llama 2-Chat 13B”. In: arXiv.2311.00117 (2024). doi: 10.48550/arXiv.2311.00117. url: https://doi.org/10.48550/arXiv. 2311.00117

2024 doi arXiv
[78]

Ethics governance development: The case of the Menlo Report

Megan Finn and Katie Shilton. “Ethics governance development: The case of the Menlo Report”. In: Social Studies of Sci- ence (2023). url: https://doi.org/10. 1177/03063127231151708

2023
[79]

Searchlights across the black box: Trade secrecy versus access to information

Katarina Foss-Solbrekk. “Searchlights across the black box: Trade secrecy versus access to information”. In: Com- puter Law and Security Review 50 (2023). doi: 10.1016/j.clsr.2023.105811. url: https://doi.org/10.1016/j.clsr.2023. 105811

2023
[80]

The tech industry can’t agree on what open-source AI means

Edd Gent. The tech industry can’t agree on what open-source AI means. That’s a problem. permalink: https://perma.cc/ P4A7-879Q. MIT Technology Review, 2024. url: https://www.technologyreview. com/2024/03/25/1090111/tech-industry- open-source-ai-definition-problem/

2024
[81]

Not all ‘open source’ AI models are actually open: here’s a rank- ing

Elizabeth Gibney. Not all ‘open source’ AI models are actually open: here’s a rank- ing. permalink: https://perma.cc/27TU- ETEK. 2024. url: https://www.nature.com/ articles/d41586-024-02012-5

2024
[82]

Dynamic Capability and Open-Source Strategy in the Age of Digital Transformation

Nobuyuki Fukawa, Yanzhi Zhang, and Sunil Erevelles. “Dynamic Capability and Open-Source Strategy in the Age of Digital Transformation”. In: Journal of Open Innovation: Technology, Market, and Complexity 7 (3 2021), p. 175. doi: 10.3390/joitmc7030175. url: https:// doi.org/10.3...

2021 doi
[83]

Public Procurement Transparency and its Potential to Reduce Corruption in Low-Income Countries

Salome Girgvliani. “Public Procurement Transparency and its Potential to Reduce Corruption in Low-Income Countries”. In: CSIS - Center for Strategic and Interna- tional Studies (2023). url: https://www. csis.org/blogs/development-dispatch/ public-procurement-transparency-and- ...

2023
[84]

Quantifying the Pressure of Legal Risks on Third-party Vulnerability Research

Alexander Gamero-Garrido, Stefan Savage, Kirill Levchenko, and Alex C. Snoeren. “Quantifying the Pressure of Legal Risks on Third-party Vulnerability Research”. In: ACM Conf. on Computers and Com- munications Security (2017), pp. 1501–

2017
[85]

Computational approaches to Explainable Artiﬁcial Intelli- gence: Advances in theory, applications and trends

J.M. G´ orriz, I. ´Alvarez-Ill´ an, A. ´Alvarez- Marquina, J.E. Arco, M. Atzmueller, F. Mallarini, E. Barakova, G. Bologna, P. Bonomini, G. Castellanos-Dominguez, D. Castillo-Barnes, S.B. Cho, R. Con- treras, J.M. Cuadra, E. Dom ´ ınguez, F. Dom ´ ınguez-Mateos, R.J. Duro, D. ...

2023
[86]

Machine Un- learning via Simulated Oracle Matching

Shivam Garg, Kristian G Georgiev, Andrew Ilyas, Sung Min Park, Aleksander Rinberg Roy adn Madry, and Seth Neel. “Machine Un- learning via Simulated Oracle Matching”. In: GenLaw 2024. 2024

2024
[88]

A global taxonomy of interpretable AI: unifying the terminology for the techni- cal and social sciences

Mara Graziani, Lidia Dutkiewicz, Davide Calvaresi, Jos´ e Pereira Amorim, Katerina Yordanova, Mor Vered, Rahul Nair, Pe- dro Henriques Abreu, Tobias Blanke, Vale- ria Pulignano, John O. Prior, Lode Lauwaert, Wessel Reijers, Adrien Depeursinge, Vin- cent Andrearczyk, and Hennin...

2022
[89]

Ap- ple, Nvidia, Anthropic Used Thousands of Swiped YouTube Videos to Train AI

Annie Gilbertson and Alex Reisner. Ap- ple, Nvidia, Anthropic Used Thousands of Swiped YouTube Videos to Train AI . Proof News, Wired. permalink: https://perma. cc/3F86-P3YR. 2024. url: https://www. proofnews.org/apple-nvidia-anthropic- used-thousands-of-swiped-youtube- videos...

2024
[90]

NIST Asks Public to Help Future-Proof Elec- tronic Information

NIST Cryptographic Technology Group. NIST Asks Public to Help Future-Proof Elec- tronic Information . permalink: https:// perma.cc/8DH6-YDXM. 2016. url: https:// www.nist.gov/news-events/news/2016/ 12/nist-asks-public-help-future- proof-electronic-information

2016
[91]

Problems of monetary management: the U.K. experience

Charles Goodhart. “Problems of monetary management: the U.K. experience”. In: Papers in monetary economics (1975), pp. 1–20

1975
[92]

Black box algorithms and the rights of individuals: no easy solution to the

Jarek Gryz and Marcin Rojszczak. “Black box algorithms and the rights of individuals: no easy solution to the ”explainability” prob- lem”. In: Internet Policy Review 10 (2 2021). doi: 10.14763/2021.2.1564

2021 doi
[93]

Towards Learning from Losing Aaron Swartz

Jennifer Granick. Towards Learning from Losing Aaron Swartz . Center for Inter- net and Security at Stanford Law School. permalink: https://perma.cc/53E5-6RDQ. url: https://cyberlaw.stanford.edu/ blog/2013/01/towards-learning-losing- aaron-swartz

2013
[94]

Community Norms as Self-Regulation of Generative AI in Creative Industries

Peter Hall, Betty Li Hou, and Falaah Arif Khan. “Community Norms as Self-Regulation of Generative AI in Creative Industries”. In: GenLaw 2024. 2024

2024
[95]

Interpreting Black-Box Models: A Review on Explainable Artiﬁcial Intelligence

Vikas Hassija, Vinay Chamola, Atmesh Ma- hapatra, Abhinandan Singal, Divyansh Goel, Kaizhu Huang, Simone Scardapane, Indro Spinelli, Mufti Mahmud, and Amir Hussain. “Interpreting Black-Box Models: A Review on Explainable Artiﬁcial Intelligence”. In: Cogni- tive Computation 16 ...

2023
[96]

Fanstastic Copyrighted Beasts and How (Not) to Generate Them

Luxi He, Yangsibo Huang, Weijia Shi, Ting- hao Xie, Haotian Liu, Yue Wang, Luke Zettle- moyer, Chiyuan Zhang, Danqi Chen, and Peter Henderson. “Fanstastic Copyrighted Beasts and How (Not) to Generate Them”. In: GenLaw 2024. 2024

2024
[97]

OLMo: Accelerating the Science of Language Models

Dirk Groeneveld et al. “OLMo: Accelerating the Science of Language Models”. In: Compu- tation and Language (2024). url: https:// arxiv.org/pdf/2402.00838

2024 arXiv
[98]

An Overview of Catas- trophic AI Risks

Dan Hendrycks, Mantas Mazeika, and Thomas Woodside. “An Overview of Catas- trophic AI Risks”. In: arXiv (2023). url: https://arxiv.org/abs/2306.12001

2023 arXiv
[99]

Post-Quantum Cryptography Minimum Acceptability Requirements

NIST Cryptographic Technology Group. Post-Quantum Cryptography Minimum Acceptability Requirements . permalink: https://perma.cc/5Z3K-TANL. 2017. url: https://csrc.nist.gov/projects/post- quantum-cryptography/post-quantum- cryptography-standardization/minimum- acceptability-req...

2017
[100]

Expe- riences with improving the transparency of AI models and services

M Hind, S Houde, J Martino, A Mojislovic, D Piorkowski, and J et al. Richard. “Expe- riences with improving the transparency of AI models and services”. In: Conf. on Hu- man Factors in Computing Systems (2020), pp. 1–8. doi: 10.1145/3334480.3383051. url: https://dl.acm.org/doi...

2020
[101]

Privacy, accuracy, and model fairness trade-oﬀs in federated learning

Xiuting Gu, Zhu Tianqing, Jie Li, Tao Zhang, Wei Ren, and Kim-Kwang Ray- mond Choo. “Privacy, accuracy, and model fairness trade-oﬀs in federated learning”. In: Computers and Security 122 (2022). doi: 10.1016/j.cose.2022.102907. url: https://doi.org/10.1016/j.cose.2022. 102907

2022
[102]

The US physics community is not done working on trust

Frances Houle, Kate Kirby, Laura Greene, and Michael Marder. The US physics community is not done working on trust . perma- link: https://perma.cc/V9U4-33ZY. MIT Technology Review, 2024. url: https:// www.technologyreview.com/2024/07/31/ 1095425/the-us-physics-community- is-no...

2024
[103]

Software Architect’s Handbook

Joseph Ingeno. Software Architect’s Handbook. Packt Publishing, 2018. url: https://www. oreilly.com/library/view/software- architects-handbook/9781788624060/ 1e2ac7d1-3f9a-4a4f-ab78-f7f9f5b9e8a6. xhtml

2018
[104]

The Open Source AI Deﬁnition - https://hackmd.io/@opensourceinitiative/osaid- 0-0-8

Open Source Initiative. The Open Source AI Deﬁnition - https://hackmd.io/@opensourceinitiative/osaid- 0-0-8. permalink: https://perma.cc/ 9A3Z-JTVR. 2024. url: https://hackmd.io/ @opensourceinitiative/osaid-0-0-8

2024
[105]

AI companies promised to self-regulate one year ago

Melissa Heikkil¨ a. AI companies promised to self-regulate one year ago. What’s changed? permalink: https://perma.cc/4ACF- DGBJ. MIT Technology Review, 2024. url: https://www.technologyreview. com/2024/07/22/1095193/ai-companies- promised-the-white-house-to-self- regulate-one-...

2024
[106]

Encryption: A Matter of Human Rights

Amnesty International. Encryption: A Matter of Human Rights . 2016. url: https://www. amnesty.org/en/wp-content/uploads/ 2021/05/POL4036822016ENGLISH.pdf

2016
[107]

’Anonymised’ data can never be totally anonymous, says study

Alex Hern. ’Anonymised’ data can never be totally anonymous, says study . perma- link: https://perma.cc/9W9X-LHGZ. 2019. url: https://www.theguardian.com/ technology/2019/jul/23/anonymised- data-never-be-anonymous-enough-study- finds. 20

2019
[108]

Crypto Wars: The Fight for Pri- vacy in the Digital Age

Craig Jarvis. Crypto Wars: The Fight for Pri- vacy in the Digital Age . CRC Press, 2020. isbn: 9780367642488

2020
[109]

Regulating AI: Getting the Balance Right

Kartik Hosanagar. “Regulating AI: Getting the Balance Right”. In: Knowledge at Whar- ton (2024). permalink: https://perma. cc/XWQ9-YH6X. url: https://knowledge. wharton.upenn.edu/article/regulating- ai-getting-the-balance-right/

2024
[110]

Highly Accurate Protein Structure Prediction with AlphaFold

John Jumper, Alexander Pritzel Richard Evans, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvu- nakool, Russ Bates, Augustin ˇZ ´ ıdek, Anna Potapenko, Alex Bridgland, Clemens Meyer, Simon A. A. Kohl, Andrew J. Ballard, Andrew Cowie, Bernardino Romera-Paredes, ...

2021
[111]

The Codebreakers: The Com- prehensive History of Secret Communication from Ancient Times to the Internet

David Kahn. The Codebreakers: The Com- prehensive History of Secret Communication from Ancient Times to the Internet . Scribner,
[112]

Introduc- tion to Modern Cryptography

Jonathan Katz and Yehuda Lindell. Introduc- tion to Modern Cryptography . Second. CRC Press, 2015

2015
[113]

The Open Source Def- inition - https://opensource.org/osd

Open Source Initiative. The Open Source Def- inition - https://opensource.org/osd . perma- link: https://perma.cc/T4MB-97MR. 2024. url: https://opensource.org/osd

2024
[114]

La Cryptographie Mil- itaire

Auguste Kerckhoﬀs. “La Cryptographie Mil- itaire”. In: Journal des sciences militaires . Vol. 9. 1883, pp. 5–38, 161–191
[115]

Methods for Adapting Large Language Models

Aditya Jain, Amir Maleki, and Nathalie Saade. Methods for Adapting Large Language Models. Tech. rep. permalink: https:// perma.cc/WC76-MBL7. Meta AI, 2024. url: https://ai.meta.com/blog/adapting- large-language-models-llms/

2024
[116]

Secrecy and Unaccountabil- ity: Trade Secrets in Our Public Infrastruc- ture

David S. Levine. “Secrecy and Unaccountabil- ity: Trade Secrets in Our Public Infrastruc- ture”. In: Florida Law Review 59 (2006). url: https://ssrn.com/abstract=900929. 21

2006
[117]

Computer Security Ver- sus the Public’s Right to Know

Douglas W. Jones. “Computer Security Ver- sus the Public’s Right to Know”. In: Computers, Freedom and Privacy (2007). url: https://homepage.divms.uiowa.edu/ ~jones/voting/cfp2007.pdf

2007
[118]

Explainable AI: A Review of Machine Learning Interpretabil- ity Methods

Pantelis Linardatos, Vasilis Papastefanopou- los, and Sotiris Kotsiantis. “Explainable AI: A Review of Machine Learning Interpretabil- ity Methods”. In: Entropy (2020). doi: 10.3390/e23010018

2020 doi
[119]

The Mythos of Model Interpretability

Zachary C. Lipton. “The Mythos of Model Interpretability”. In: eprint arXiv:1606.03490 (2016). url: https://arxiv.org/abs/1606. 03490

2016 arXiv
[120]

Digital rights man- agement for content distribution

Qiong Liu, Reihanah Safavi-Naini, and Nicholas Paul Sheppard. “Digital rights man- agement for content distribution”. In: ACM ACSW Frontiers ‘03. ACM ACSW Frontiers, 2003, pp. 49–58

2003
[121]

October 26, 1992: Software Glitch Cripples Ambulance Service

Tony Long. October 26, 1992: Software Glitch Cripples Ambulance Service . perma- link: https://perma.cc/ZE6Q-FCJ2. 2009. url: https://www.wired.com/2009/ 10/1026london-ambulance-computer- meltdown/

1992
[122]

Biosecurity Culture, Com- puter Security Culture

Jeﬀ Kaufman. Biosecurity Culture, Com- puter Security Culture . permalink: https://perma.cc/5AG3-JMT3. 2023. url: https://forum.effectivealtruism.org/ posts/PTtZWBAKgrrnZj73n/biosecurity- culture-computer-security-culture

2023
[123]

Black-Box vs. White-Box: Understanding Their Advantages and Weaknesses From a Practical Point of View

Octavio Loyola-Gonz´ alez. “Black-Box vs. White-Box: Understanding Their Advantages and Weaknesses From a Practical Point of View”. In: IEEE Access 7 (1 2019). doi: 10.1109/ACCESS.2019.2949286. url: https://ieeexplore.ieee.org/document/ 8882211

2019
[124]

Drawbacks of Artiﬁcial Intelligence and Their Potential Solutions in the Healthcare Sector

Bangul Khan, Hajira Fatima, Ayatullah Qureshi, Sanjay Kumar, Abdul Hanan, Jawad Hussain, and Saad Abdullah. “Drawbacks of Artiﬁcial Intelligence and Their Potential Solutions in the Healthcare Sector”. In: Biomedical Materials and Devices (2023). doi: 10.1007/s44174-023-00063-...

2023 doi
[125]

Thoughts on open source AI

Sam Marks. Thoughts on open source AI . permalink: https://perma.cc/292X-QZJK
[126]

Rethinking open source generative AI: open-washing and the EU AI Act

Andreas Liesenfeld and Mark Dingemanse. “Rethinking open source generative AI: open-washing and the EU AI Act”. In: F AccT 2024 (2024), pp. 1774–1787. doi: 10.1145/3630106.3659005. url: https:// doi.org/10.1145/3630106.3659005

2024
[127]

A multi-objective op- timization design to generate surrogate machine learning models in explainable artiﬁcial intelligence applications

Wellington Rodrigo Monteiro and Gilberto Reynoso-Meza. “A multi-objective op- timization design to generate surrogate machine learning models in explainable artiﬁcial intelligence applications”. In: EURO Journal on Decision Processes 11 (2023). doi: 10.1016/j.ejdp.2023.100040....

2023
[128]

Vulnerability Disclosure Considered Stress- ful

Giovane C. M. Moura and John Heidemann. “Vulnerability Disclosure Considered Stress- ful”. In: SIGCOMM Computer Communica- tion Review . Vol. 53. ACM SIGCOMM Com- puter Communication Review, 2023, pp. 2–10. doi: 10.1145/3610381.3610383

2023
[129]

Robust De-anonymization of Large Sparse Datasets

Arvind Narayanan and Vitaly Shmatikov. “Robust De-anonymization of Large Sparse Datasets”. In: 2008 IEEE Symposium on Security and Privacy . IEEE Sympo- sium on Security and Privacy, 2008. doi: 10.1109/SP.2008.33

2008 doi
[130]

Scalable Extraction of Training Data from (Production) Language Models

Milad Nasr, Nicholas Carlini, Jonathan Hayase, Matthew Jagielski, A. Feder Cooper, Daphne Ippolito, Christopher A. Choquette- Choo, Eric Wallace, Florian Tram` er, and Katherine Lee. “Scalable Extraction of Training Data from (Production) Language Models”. In: arXiv.2311.17035...

2023 doi arXiv
[131]

A Safe Harbor for AI Evaluation and Red Teaming

Shayne Longpre, Sayash Kapoor, Kevin Kly- man, Ashwin Ramaswami, Rishi Bommasani, Borhane Blili-Hamelin, Yangsibo Huang, Aviya Skowron, Zheng-Xin Yong, Suhas Kotha, Yi Zeng, Weiyan Shi, Xianjun Yang, Reid Southen, Alexander Robey, Patrick Chao, Diyi Yang, Ruoxi Jia, Daniel Kan...

2024 doi arXiv
[132]

United States

Nathan Van Buren v. United States . url: https://www.eff.org/files/2020/07/ 08/19-783_eff_security_researchers_ amici_brief_.pdf. 22

2020
[133]

When is Technology Too Dangerous to Release to the Public?

Aaron Mak. “When is Technology Too Dangerous to Release to the Public?” In: Slate (2019). permalink: https://perma. cc/N9ZL-Z9XB. url: https://slate.com/ technology/2019/02/openai-gpt2-text- generating-algorithm-ai-dangerous. html

2019
[135]

Accountability in ar- tiﬁcial intelligence: what it is and how it works

Claudio Novelli, Mariarosaria Taddeo, and Luciano Floridi. “Accountability in ar- tiﬁcial intelligence: what it is and how it works”. In: AI and Society (2023). doi: 10.1007/s00146-023-01635-y . url: https://doi.org/10.1007/s00146-023- 01635-y

2023 doi
[136]

Security by Obscurity

Rebecca T. Mercuri and Peter G. Neumann. “Security by Obscurity”. In: Communications of the ACM 46 (2003)

2003
[137]

Broken Promises of Privacy: Responding to the surprising failure of anonymization

Paul Ohm. “Broken Promises of Privacy: Responding to the surprising failure of anonymization”. In: UCLA Law Review 57 (2009). url: https://papers.ssrn.com/ sol3/papers.cfm?abstract_id=1450006

2009
[138]

A Re- searcher’s Guide to Some Legal Risks of Security Research

Sunoo Park and Kendra Albert. A Re- searcher’s Guide to Some Legal Risks of Security Research . permalink: https:// perma.cc/27TU-ETEK. 2024. url: https:// clinic.cyber.harvard.edu/wp-content/ uploads/2024/08/Security-Researchers- Guide-8-2-24.pdf

2024
[139]

Bring- ing Transparency to National Security Uses of Artiﬁcial Intelligence

Faiza Patel and Patrick C. Toomey. “Bring- ing Transparency to National Security Uses of Artiﬁcial Intelligence”. In: justse- curity.org (2024). permalink: https:// perma.cc/SL93-RAK3. url: https:// www.justsecurity.org/94113/bringing- transparency-to-national-security- uses-o...

2024
[140]

Beneﬁts-Risk Frameworks: Implementation by Industry

Nancy J. Pire-Smerkanich. “Beneﬁts-Risk Frameworks: Implementation by Industry”. PhD thesis. University of Southern California, 2016

2016
[141]

Comprehensive Privacy Analysis of Deep Learning: Passive and Active White-box Inference Attacks against Centralized and Federated Learning

Milad Nasr, Reza Shokri, and Amir Houmansadr. “Comprehensive Privacy Analysis of Deep Learning: Passive and Active White-box Inference Attacks against Centralized and Federated Learning”. In: 2019 IEEE Symposium on Security and Privacy. IEEE Symposium on Security and Privacy, ...

2019
[142]

Nearly 400 car crashes in 11 months involved automated tech, com- panies tell regulators

The Associated Press. Nearly 400 car crashes in 11 months involved automated tech, com- panies tell regulators . permalink: https:// perma.cc/UZD8-8GK7. 2022. url: https:// www.npr.org/2022/06/15/1105252793/ nearly-400-car-crashes-in-11-months- involved-automated-tech-companie...

2022
[143]

Taxonomy of Trustwor- thiness for Artiﬁcial Intelligence: Connecting Properties of Trustworthiness with Risk Man- agement and the AI Lifecycle

Jessica Newman. “Taxonomy of Trustwor- thiness for Artiﬁcial Intelligence: Connecting Properties of Trustworthiness with Risk Man- agement and the AI Lifecycle”. In: CLTC White Paper Series (2023)

2023
[144]

Raimondo and Laurie E

Gina M. Raimondo and Laurie E. Locas- cio. Artiﬁcial Intelligence Risk Management Framework (AI RMF 1.0) . Tech. rep. Na- tional Institute of Standards and Technology,
[145]

AI in health and medicine

Pranav Rajpurkar, Emma Chen, Oishi Banerjee, and Eric J. Topol. “AI in health and medicine”. In: Nature Medicine 28 (2022), pp. 31–38. doi: 10.1038/s41591-021-01614-0

2022 doi
[146]

No, hashing still doesn’t make your data anonymous

Staﬀ in the Oﬃce of Technology. No, hashing still doesn’t make your data anonymous . permalink: https://perma.cc/HHA2-HJFX
[147]

url: https://www.ftc.gov/policy/ advocacy-research/tech-at-ftc/2024/ 07/no-hashing-still-doesnt-make-your- data-anonymous

2024
[148]

Eric S. Raymond. The Cathedral and the Bazaar: Musings on Linux and Open Source by an Accidental Revolutionary . O’Reilly and Associates, 1999. isbn: 1-565-92724-9. 23

1999
[149]

Renieris, David Kiron, and Steven Mills

Elizabeth M. Renieris, David Kiron, and Steven Mills. Organizations Face Challenges in Timely Compliance With the EU AI Act . permalink: https://perma.cc/247V-YVL7. MIT Sloan Management Review, 2024. url: https://sloanreview.mit.edu/article/ organizations-face-challenges-in- t...

2024
[150]

The Moral Character of Cryptographic Work

Phillip Rogaway. The Moral Character of Cryptographic Work. Invited Talk - AsiaCrypt
[151]

Procuring Algorithmic Transparency

Elizabeth A. Rowe. “Procuring Algorithmic Transparency”. In: Alabama Law Review 74 (2 2022). url: https://papers.ssrn.com/ sol3/papers.cfm?abstract_id=4044178

2022
[152]

Challenges and eﬀorts in managing AI trustworthiness risks: a state of knowledge

Nineta Polemi, Isabel Pra¸ ca, Kitty Kioskli, and Adrien B´ ecue. “Challenges and eﬀorts in managing AI trustworthiness risks: a state of knowledge”. In: Frontiers in Big Data (2024). doi: 10.3389/fdata.2024.1381163. url: https://doi.org/10.3389/fdata.2024. 1381163

2024
[153]

Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead

Cynthia Rudin. “Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead”. In: Nature Machine Intelligence (2022). doi: 10.1038/s42256-019-0048-x . url: https://doi.org/10.1038/s42256- 019-0048-x

2022 doi
[154]

Euro- pean Comission

Proposal for a Regulation of the European Parliament and of the Council: Laying Down Harmonised Rules on Artiﬁcial Intelligence (Artiﬁcial Intelligence Act) and Amending Certain Union Legislastive Acts . Euro- pean Comission. 2021. url: https://eur- lex.europa.eu/legal-content...

2021
[155]

Beyond Fear: Thinking Sen- sibly About Security in an Uncertain World

Bruce Schneier. Beyond Fear: Thinking Sen- sibly About Security in an Uncertain World . Springer, 2003. doi: 10.1007/b97547

2003 doi
[156]

url: https://doi.org/10.6028/NIST.AI.100-1

doi: 10.6028/NIST.AI.100-1. url: https://doi.org/10.6028/NIST.AI.100-1

doi
[157]

The Non-Security of Se- crecy

Bruce Schneier. “The Non-Security of Se- crecy”. In: Communications of the ACM 47 (2004)

2004
[158]

Red-Teaming the Stable Diﬀusion Safety Filter

Javier Rando, Daniel Paleka, David Lind- ner, Lennart Heim, and Florian Tram` er. “Red-Teaming the Stable Diﬀusion Safety Filter”. In: Artiﬁcial Intelligence (2022). doi: 10.48550/arXiv.2210.04610. url: https://doi.org/10.48550/arXiv.2210. 04610

2022 doi arXiv
[159]

Operation Charlie: Hack- ing the MBTA CharlieCard from 2008 to Present

Bobby Rauch. “Operation Charlie: Hack- ing the MBTA CharlieCard from 2008 to Present”. In: Bobbyrsec-Medium (2022). permalink: https://perma.cc/82SP-FLSE. url: https://medium.com/@bobbyrsec/ operation-charlie-hacking-the-mbta- charliecard-from-2008-to-present- 24ea9f0aaa38

2022
[160]

Communication Theory of Secrecy Systems

Claude E. Shannon. “Communication Theory of Secrecy Systems”. In: Bell Systems Techni- cal Journal . Vol. 28. 1949, pp. 656–715. url: https://pages.cs.wisc.edu/~rist/642- spring-2014/shannon-secrecy.pdf

1949
[161]

Tragedy of the Dig- ital Commons

Chinmayi Sharma. “Tragedy of the Dig- ital Commons”. In: 101 North Car- olina Law Review 1129 . 2022. doi: 10.2139/ssrn.4245266. url: https:// papers.ssrn.com/sol3/papers.cfm? abstract_id=4245266

2022 doi
[162]

Don’t Be Afraid to Code in the Open: Here’s How to Do It Securely

Anna Shipman. Don’t Be Afraid to Code in the Open: Here’s How to Do It Securely . UK Technology Blog. permalink: https:// perma.cc/3TBD-QPBC. 2019. url: https:// technology.blog.gov.uk/2017/09/27/ dont-%20be-afraid-to-code-in-the- open-heres-how-to-do-it-securely

2019
[163]

Dual-use open source secu- rity software in organizations - Dilemma: help or hinder?

Mario Silic. “Dual-use open source secu- rity software in organizations - Dilemma: help or hinder?” In: Computers and Secu- rity. Vol. 39. Elsevier, 2013, pp. 386–395. doi: 10.1016/j.cose.2013.09.003

2013 doi
[164]

Factors Associated With Pub- lic Trust in Pharmaceutical Manufacturers

Yashaswini Singh, Matthew D. Eisenberg, and Neeraj Sood. “Factors Associated With Pub- lic Trust in Pharmaceutical Manufacturers”. In: JAMA Network Open 6 (3 2023). doi: 10.1001/jamanetworkopen.2023.3002. url: https://doi.org/10.1001/ jamanetworkopen.2023.3002

2023
[165]

Anonymization and Risk

Ira S. Rubinstein and Woodrow Hartzog. “Anonymization and Risk”. In: Washing- ton Law Review 91 (2016). url: https:// digitalcommons.law.uw.edu/wlr/vol91/ iss2/18

2016
[166]

The Fundamental Limits of Least- Privilege Learning

Theresa Stadler, Bogdan Kulynych, Michael C. Gastpar, Nicolas Papernot, and Carmela Troncoso. “The Fundamental Limits of Least- Privilege Learning”. In: arXiv (2024). url: https://arxiv.org/pdf/2402.12235

2024 arXiv
[167]

AI as a challenge for legal regulation - the scope of application of the artiﬁcial intelligence act proposal

Hannah Ruschemeier. “AI as a challenge for legal regulation - the scope of application of the artiﬁcial intelligence act proposal”. In: ERA Forum (2023), pp. 361–376. doi: 10.1007/s12027-022-00725-6 . url: https://doi.org/10.1007/s12027-022- 00725-6

2023 doi
[168]

Protections for Human Sub- jects in Research: Old Models, New Needs?

Laura Stark. “Protections for Human Sub- jects in Research: Old Models, New Needs?” In: MIT Case Studies in Social and Eth- ical Responsibilities of Computing (2022). url: https://mit-serc.pubpub.org/ pub/protections-for-human-subjects/ release/1

2022
[169]

Open Source and Security

Bruce Schneier. Open Source and Security
[170]

Replacing Personally- Identifying Information in Medical Records, the Scrub System

Latanya Sweeney. Replacing Personally- Identifying Information in Medical Records, the Scrub System . 1996. url: https:// dataprivacylab.org/projects/scrub/ paper1.pdf

1996
[171]

Simple Demographics Often Identify People Uniquely

Latanya Sweeney. “Simple Demographics Often Identify People Uniquely”. In: Carnegie Mellon University, Data Privacy Working Pa- per (2000). url: https://dataprivacylab. org/projects/identifiability/paper1. pdf

2000
[172]

Trustworty AI Means Public AI [Last Word]

Bruce Schneier. “Trustworty AI Means Public AI [Last Word]”. In: IEEE Secu- rity and Privacy 21 (2023), pp. 95–96. doi: 10.1109/MSEC.2023.3301262. url: https://doi.ieeecomputersociety.org/ 10.1109/MSEC.2023.3301262

2023
[173]

The Belmont Report - Ethical Principles and Guidelines for the Protection of Human Subjects of Research

Oﬃce of the Secretary. The Belmont Report - Ethical Principles and Guidelines for the Protection of Human Subjects of Research . The National Commission for the Protec- tion of Human Subjects of Biomedical and Behavioral Research, U.S.A. 1979. url: https://www.hhs.gov/ohrp/reg...

1979
[174]

P=NP? Not exactly, but here are some research questions from the Oﬃce of Technology

FTC Staﬀ - Oﬃce of Technology. P=NP? Not exactly, but here are some research questions from the Oﬃce of Technology . U.S. Federal Trade Commission. permalink: https:// perma.cc/4ELG-H7XX. 2024. url: https:// www.ftc.gov/policy/advocacy-research/ tech-at-ftc/2024/05/p-np-not-ex...

2024
[175]

Technolo- gies Underlying Weapons of Mass Destruc- tion

Oﬃce of Technology Assessment. Technolo- gies Underlying Weapons of Mass Destruc- tion. U.S. Government Printing Oﬃce, 1993. isbn: 0160430224. url: https://ota.fas. org/reports/9344.pdf

1993
[176]

OpenAI says voice cloning tool too risky to release, cites fears of “mis- use

Shannon Thaler. “OpenAI says voice cloning tool too risky to release, cites fears of “mis- use” in 2024 election”. In: NY Post (2024). permalink: https://perma.cc/5M6W- 7WKP. url: https://nypost.com/2024/04/ 01/business/openai-says-its-voice- cloning-tool-too-risky-to-release/

2024
[177]

Microsoft Corporation et al., No

The New York Times Company v. Microsoft Corporation et al., No. 1:23-cv-11195 ( S.D. N.Y. 2023). url: https://nytco-assets. nytimes.com/2023/12/NYT_Complaint_ Dec2023.pdf

2023
[178]

UN General Assembly

The right to privacy in the digital age- Re- port of the Oﬃce of the United Nations High Commissioner for Human Rights . UN General Assembly. 2022

2022
[179]

The Ballot is Busted Before the Blockchain: A Security Analysis of Voatz, the First Internet Voting Application Used in U.S. Federal Elections

Michael A. Specter, James Koppel, and Daniel J. Weitzner. “The Ballot is Busted Before the Blockchain: A Security Analysis of Voatz, the First Internet Voting Application Used in U.S. Federal Elections”. In: 29th USENIX Security Symposium, USENIX Security 2020, August 12-14, 2...

2020
[180]

Top AI researchers say Ope- nAI, Meta and more hinder independent evaluations

Nitasha Tiku. Top AI researchers say Ope- nAI, Meta and more hinder independent evaluations. Washington Post. perma- link: https://perma.cc/N7B6-MUVH. 2024. url: https://www.washingtonpost.com/ technology/2024/03/05/ai-research- letter-openai-meta-midjourney/

2024
[181]

Behind Closed Doors: IRBs and the Making of Ethical Research

Laura Stark. Behind Closed Doors: IRBs and the Making of Ethical Research . University of Chicago Press, 2011. url: https://press. uchicago.edu/ucp/books/book/chicago/ B/bo12182576.html

2011
[182]

Llama 2: Open Foun- dation and Fine-Tuned Chat Models

Hugo Touvron et al. “Llama 2: Open Foun- dation and Fine-Tuned Chat Models”. In: Computation and Language (2023). doi: 10.48550/arXiv.2307.09288. url: https://doi.org/10.48550/arXiv.2307. 09288

2023 doi arXiv
[183]

A Frame- work for Understanding Sources of Harm throughout the Machine Learning Life Cycle

Harini Suresh and John Guttag. “A Frame- work for Understanding Sources of Harm throughout the Machine Learning Life Cycle”. In: Equity and Access in Algorithms Mech- anisms, and Optimization . ACM, 2021. doi: 10.1145/3465416.3483305

2021
[184]

Regulation (EU) 2024/1689 of the European Parliament and of the Council

European Union. “Regulation (EU) 2024/1689 of the European Parliament and of the Council”. In: Oﬃcial Journal of the European Union (2024). url: http:// data.europa.eu/eli/reg/2024/1689/oj

2024
[185]

Cybersecurity and Moral Haz- ard

Jeﬀrey Vagle. “Cybersecurity and Moral Haz- ard”. In: Stanford Technology Law Review (2020). doi: 10.2139/ssrn.3055231

2020 doi
[186]

The pace of artiﬁcial inteligence innovations: Speed, talent, and trial-and- error

Xuli Tang, Xin Li, Ying Ding, Min Song, and Yi Bu. “The pace of artiﬁcial inteligence innovations: Speed, talent, and trial-and- error”. In: Journal of Informetrics 14 (4 2020). doi: 10.1016/j.joi.2020.101094. url: https://doi.org/10.1016/j.joi. 2020.101094

2020
[187]

On Open- Weights Foundation Models

FTC Staﬀ - Oﬃce of Technology. On Open- Weights Foundation Models . U.S. Federal Trade Commission. permalink: https:// perma.cc/P5TJ-E5JX. 2024. url: https:// www.ftc.gov/policy/advocacy-research/ tech-at-ftc/2024/07/open-weights- foundation-models

2024
[188]

OpenAI has published the text-generating AI it said was too dan- gerous to share

James Vincent. “OpenAI has published the text-generating AI it said was too dan- gerous to share”. In: The Verge (2019). permalink: https://perma.cc/YY32-LABJ. url: https://www.theverge.com/2019/ 11/7/20953040/openai-text-generation- ai-gpt-2-full-model-release-1-5b- parameters

2019
[189]

Jailbroken: How Does LLM Safety Training Fail?

Alexander Wei, Nika Haghtalab, and Jacob Steinhardt. “Jailbroken: How Does LLM Safety Training Fail?” In: Machine Learning (2023). doi: 10.48550/arXiv.2307.02483. url: https://doi.org/10.48550/arXiv. 2307.02483

2023 doi arXiv
[190]

Ethical and social risks of harm from language models

Laura Weidinger, John Mellor, Maribeth Rauh, Conor Griﬃn, Jonathan Uesato, Po- Sen Huang, Myra Cheng, Mia Glaese, Borja Balle, Atoosa Kasirzadeh, et al. “Ethical and social risks of harm from language models”. In: arXiv:2112.04359 (2021). doi: arXiv:2112.04359

2021 arXiv
[191]

Good Faith Researcher Code of Conduct

Tarah Wheeler. Good Faith Researcher Code of Conduct . permalink: https:// perma.cc/X5QT-3Y6N. 2022. url: https:// tarahmarie.github.io/gfcrc/

2022
[192]

The Model Openness Framework: Promoting Completeness and Openness for Repro- ducibility, Transparency, and Usability in Artiﬁcial Intelligence

Matt White, Ibrahim Haddad, Cailean Osborne, Xiao-Yang Liu Yanglet, Ahmed Abdelmonsef, and Sachin Varghese. The Model Openness Framework: Promoting Completeness and Openness for Repro- ducibility, Transparency, and Usability in Artiﬁcial Intelligence . arXiv.2403.13784

arXiv
[193]

Ethical issues in research us- ing datasets of illicit origin

Daniel R. Thomas, Sergio Pastrana, Alice Hutchings, Richard Clayton, and Alastair R. Beresford. “Ethical issues in research us- ing datasets of illicit origin”. In: IMC ’17: Proceedings of the 2017 Internet Measure- ment Conference. London, UK, 2017. doi: 10.1145/3131365.31313...

2017
[194]

permalink: https://perma.cc/ GAA6-WL6U

WIPO Guide to Trade Secrets and Inno- vation. permalink: https://perma.cc/ GAA6-WL6U. World Intellectual Property Organization. url: https://www.wipo.int/ web-publications/wipo-guide-to-trade- secrets-and-innovation/en/index.html
[195]

Code Talkers Were America’s Se- cret Weapon in World War II

Laura Tohe. Code Talkers Were America’s Se- cret Weapon in World War II . Humanities: The Magazine of the NEH. 2022. 25

2022
[196]

Diﬀerentially Private Fine-tuning of Language Models

Da Yu, Saurabh Naik, Arturs Backurs, Sivakanth Gopi, Huseyin A. Inan, Gau- tam Kamath, Janardhan Kulkarni, Yin Tat Lee, Andre Manoel, Lukas Wutschitz, Sergey Yekhanin, and Huishuai Zhang. “Diﬀerentially Private Fine-tuning of Language Models”. In: J. Priv. Conﬁdentiality 14.2 ...

2024 doi
[197]

Beyond the Black Box

Charlotte Tschider. “Beyond the Black Box”. In: Denver Law Review 98 (2021). url: https://papers.ssrn.com/sol3/papers. cfm?abstract_id=3855782#

2021
[198]

A Systematic Review of Multimedia Tools for Cybersecurity Awareness and Education

Leah Zhang-Kennedy and Sonia Chiasson. “A Systematic Review of Multimedia Tools for Cybersecurity Awareness and Education”. In: Computing Surveys (CSUR) (1 2021), pp. 1–

2021
[199]

Discretion and the quest for controlled freedom

Stavros Zouridis and Marlies Van Eckand M.A.P. Bovens. “Discretion and the quest for controlled freedom”. In: Palgrave Macmillan,
[200]

Speech at the European Commission workshop on ”Competition in Virtual Worlds and Generative AI”

Margrethe Vestager. Speech at the European Commission workshop on ”Competition in Virtual Worlds and Generative AI” . perma- link: https://perma.cc/6KP7-3BT4. 2024. url: https://ec.europa.eu/commission/ presscorner/detail/en/SPEECH_24_3550

2024
[201]

Building Se- cure Software: How to Avoid Security Prob- lems the Right Way

John Viega and Gary McGraw. Building Se- cure Software: How to Avoid Security Prob- lems the Right Way . Addison-Wesley Profes- sional, 2001. isbn: 9780672334092

2001
[207]

url: https://arxiv.org/abs/2403.13784

doi: 10.48550/arXiv.2403.13784. url: https://arxiv.org/abs/2403.13784

doi arXiv
[208]

The Ethics of Zero-Day Exploits: The NSA Meets the Trolley Car

Stephen B. Wicker. “The Ethics of Zero-Day Exploits: The NSA Meets the Trolley Car”. In: Communications of the ACM 64 (1 2021). doi: 10.1145/3393670

2021 doi
[210]

Diﬀerentially Private Post- Processing for Fair Regression

Ruicheng Xian, Qiaobo Li, Gautam Kamath, and Han Zhao. “Diﬀerentially Private Post- Processing for Fair Regression”. In: Forty-ﬁrst International Conference on Machine Learn- ing, ICML 2024, Vienna, Austria, July 21-27,

2024
[211]

url: https:// openreview.net/forum?id=JNeeRjKbuH

OpenReview.net, 2024. url: https:// openreview.net/forum?id=JNeeRjKbuH

2024
[213]

Solving the Black Box Problem: A Normative Framework for Explainable Artiﬁcial Intelligence

Carlos Zednik. “Solving the Black Box Problem: A Normative Framework for Explainable Artiﬁcial Intelligence”. In: Phi- losophy and Technology 34 (2019), pp. 265–

2019
[216]

doi: 10.1145/3427920. 26

doi
[218]

Automated Discretion, pp

Chap. Automated Discretion, pp. 313– 329
[219]

Open Source AI Is the Path Forward

Mark Zuckerberg. Open Source AI Is the Path Forward. Meta Newsroom. permalink: https://perma.cc/755A-Y4ET. 2024. url: https://about.fb.com/news/2024/07/ open-source-ai-is-the-path-forward/ . 27

2024
[288]

url: https://doi.org/10.1007/s13347- 019-00382-7

doi: 10.1007/s13347-019-00382-7 . url: https://doi.org/10.1007/s13347- 019-00382-7

doi
[589]

doi: 10.1038/s41586-021-03819-2

doi
[1331]

url: https://link.springer.com/ article/10.1007/s00146-021-01326-6

doi
[1513]

url: https://dl.acm.org/doi/10.1145/ 3133956.3134047

doi: 10.1145/3133956.3134047. url: https://dl.acm.org/doi/10.1145/ 3133956.3134047
[1999]

url: https://www.schneier.com/ crypto-gram/archives/1999/0915.html# OpenSourceandSecurity

1999
[2015]

url: https://web.cs.ucdavis

2015. url: https://web.cs.ucdavis. edu/~rogaway/papers/moral-fn.pdf

2015
[2019]

url: https://ai.google/static/ documents/perspectives-on-issues-in- ai-governance.pdf
[2023]

url: https://www.lesswrong.com/ posts/WLYBy5Cus4oRFY3mu/thoughts-on- open-source-ai#fnlndvavl4r2a
[2024]

url: https://certcc.github.io/ CERT-Guide-to-CVD/topics/phases/ deployment/
[3504]

url: https://link.springer.com/ article/10.1007/s10462-022-10256-8

doi

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.

[1] [1]

Artiﬁcial Intelligence and Jobs — Evidence from Online Vacancies

Daron Acemoglu, David Autor, Jonathon Hazell, and Pascual Restrepo. “Artiﬁcial Intelligence and Jobs — Evidence from Online Vacancies”. In: Journal of Labor Economics 40 (1) (2022), pp. 293–340. doi: 10.1086/718327. url: https:// economics.mit.edu/sites/default/ files/publications/AI%20and%20Jobs %20-%20Evidence%20from%20Online %20Vacancies.pdf

doi:10.1086/718327 2022

[2] [2]

RSA Factor- ing Challenge Announcement

RSA Challenge Administrator. RSA Factor- ing Challenge Announcement . permalink: https://perma.cc/Z3W7-FCDU. 1991. url: https://groups.google.com/u/1/g/sci. crypt/c/AA7M9qWWx3w/m/EkrsR69CDqIJ? pli=1

1991

[3] [3]

Infor- mation Sharing

America’s Cyber Defense Agency. Infor- mation Sharing . permalink: https:// perma.cc/3GP9-MD94. url: https://www. cisa.gov/topics/cyber-threats-and- advisories/information-sharing

[4] [4]

Perspectives on Issues in AI Governance

Google AI. Perspectives on Issues in AI Governance . Tech. rep. Google,

[5] [5]

Ethics Guidelines for Trustworthy AI

High-Level Expert Group on AI. Ethics Guidelines for Trustworthy AI . European Commission. 2019

2019

[6] [6]

Piloting a survey- based assessment of transparency and trustworthiness with three medical AI tools

Jana Fehr et al. “Piloting a survey- based assessment of transparency and trustworthiness with three medical AI tools”. In: Healthcare (Basel) (2022). doi: 10.3390/healthcare10101923. url: https://www.ncbi.nlm.nih.gov/pmc/ articles/PMC9601535/

doi:10.3390/healthcare10101923 2022

[7] [7]

Apple pays hackers six ﬁgures to ﬁnd bugs in its software. Then it sits on their ﬁndings

Reed Albergotti. “Apple pays hackers six ﬁgures to ﬁnd bugs in its software. Then it sits on their ﬁndings.” In: Washing- ton Post (2021). permalink: https:// perma.cc/5LTD-4YBW. url: https://www. washingtonpost.com/technology/2021/ 09/09/apple-bug-bounty/

2021

[8] [8]

The Falcon Series of 14 Open Language Models

Ebtesam Almazrouei, Hamza Alobeidli, Abdulaziz Alshamsi, Alessandro Cappelli, Ruxandra Cojocaru, M´ erouane Debbah, Etienne Goﬃnet, Daniel Hesslow, Julien Launay, Quentin Malartic, Daniele Mazzotta, Badreddine Noune, Baptiste Pannier, and Guilherme Penedo. “The Falcon Series of 14 Open Language Models”. In: Computation and Language (2023). url: https://ar...

arXiv 2023