{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:MG6H7RI5SOKDIPKDJLMOZOGFVS","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"d7652b21ddac548046c4219fa570c258bc5689316147dedc995117c6ef4ea1f3","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-05-03T17:07:20Z","title_canon_sha256":"3cf963ab7d00868573c79105b59830e14ee4fb915689f5c8bce871cf35b68a40"},"schema_version":"1.0","source":{"id":"2605.01970","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.01970","created_at":"2026-05-20T00:00:40Z"},{"alias_kind":"arxiv_version","alias_value":"2605.01970v3","created_at":"2026-05-20T00:00:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.01970","created_at":"2026-05-20T00:00:40Z"},{"alias_kind":"pith_short_12","alias_value":"MG6H7RI5SOKD","created_at":"2026-05-20T00:00:40Z"},{"alias_kind":"pith_short_16","alias_value":"MG6H7RI5SOKDIPKD","created_at":"2026-05-20T00:00:40Z"},{"alias_kind":"pith_short_8","alias_value":"MG6H7RI5","created_at":"2026-05-20T00:00:40Z"}],"graph_snapshots":[{"event_id":"sha256:a34f3fd04ab357a5c150efe73cdc2ddaf15e34d8ef9591f3df1f69f1353b0110","target":"graph","created_at":"2026-05-20T00:00:40Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Trojan Hippo achieves up to 85-100% ASR against current frontier models from OpenAI and Google, with planted memories successfully activating even after 100 benign sessions."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The evaluation assumes that the four memory backends (explicit tool memory, agentic memory, RAG, and sliding-window context) and the OpenEvolve-based adaptive red-teaming accurately represent real-world deployed agent systems and that the threat model of a single untrusted tool call is realistic for attackers."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"The paper defines and evaluates Trojan Hippo attacks on LLM agent memory, showing 85-100% success in data exfiltration across backends and reduced rates with defenses at varying utility costs."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A single untrusted tool call can plant a dormant payload in an agent's memory that later activates to exfiltrate sensitive user data."}],"snapshot_sha256":"c107e90e2b1e07816a83e1f68d4695f039ce806e17d3bc0a165280654c5ef867"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"3932c6654419eddb7dee52d0fa757737dd5f7f47e26d8aa900849bb13b8916f8"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T16:50:08.716171Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.01970/integrity.json","findings":[],"snapshot_sha256":"72d287e6f2159dbc77248144fee92a0d4ca22a2992adcb442dc6813fcde158e1","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Memory systems enable otherwise-stateless LLM agents to persist user information across sessions, but also introduce a new attack surface. We characterize the Trojan Hippo attack, a class of persistent memory attacks that operates in a more realistic threat model than prior memory poisoning work: the attacker plants a dormant payload into an agent's long-term memory via a single untrusted tool call (e.g., a crafted email), which activates only when the user later discusses sensitive topics such as finance, health, or identity, and exfiltrates high-value personal data to the attacker.\n  While a","authors_text":"Darya Kaviani, David Wagner, Debeshee Das, Florian Tram\\`er, Julien Piet, Luca Beurer-Kellner","cross_cats":["cs.AI"],"headline":"A single untrusted tool call can plant a dormant payload in an agent's memory that later activates to exfiltrate sensitive user data.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-05-03T17:07:20Z","title":"Trojan Hippo: Weaponizing Agent Memory for Data Exfiltration"},"references":{"count":104,"internal_anchors":19,"resolved_work":104,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"https://openai.com/index/scaling-ai-for- everyone/","work_id":"dc16c4c9-f5b7-4439-96da-a9e3cd7460a6","year":2026},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Trustworthy agentic ai systems: a cross-layer review of architectures, threat models, and governance strategies for real-world deployment.F1000Research, 14(905):905, 2025","work_id":"2c481df5-6b5f-4b61-ac36-e4a1d030f6e0","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Burtsev, and Evgeny Burnaev","work_id":"5e87b079-06e6-4268-83b2-f336d186debd","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Claude memory","work_id":"62c00962-89b8-40c0-9d6f-af29f0d870a2","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Import your ChatGPT history to Claude","work_id":"d55175f9-7904-4bba-a413-45dd5bb6c634","year":2026}],"snapshot_sha256":"3dfe87db60f7cc033ca973637399230ccf7477a426fe125369ced4a2f6a04884"},"source":{"id":"2605.01970","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-19T17:26:07.958042Z","id":"865f2c63-71be-4adf-bc4b-953e4f912623","model_set":{"reader":"grok-4.3"},"one_line_summary":"The paper defines and evaluates Trojan Hippo attacks on LLM agent memory, showing 85-100% success in data exfiltration across backends and reduced rates with defenses at varying utility costs.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A single untrusted tool call can plant a dormant payload in an agent's memory that later activates to exfiltrate sensitive user data.","strongest_claim":"Trojan Hippo achieves up to 85-100% ASR against current frontier models from OpenAI and Google, with planted memories successfully activating even after 100 benign sessions.","weakest_assumption":"The evaluation assumes that the four memory backends (explicit tool memory, agentic memory, RAG, and sliding-window context) and the OpenEvolve-based adaptive red-teaming accurately represent real-world deployed agent systems and that the threat model of a single untrusted tool call is realistic for attackers."}},"verdict_id":"865f2c63-71be-4adf-bc4b-953e4f912623"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:655e5425fa2bfa4a6bc890e1dda0557bb63b6963da12e07a57f11f76b793472e","target":"record","created_at":"2026-05-20T00:00:40Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"d7652b21ddac548046c4219fa570c258bc5689316147dedc995117c6ef4ea1f3","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-05-03T17:07:20Z","title_canon_sha256":"3cf963ab7d00868573c79105b59830e14ee4fb915689f5c8bce871cf35b68a40"},"schema_version":"1.0","source":{"id":"2605.01970","kind":"arxiv","version":3}},"canonical_sha256":"61bc7fc51d9394343d434ad8ecb8c5ac91bd544eacbf854aea228dde90169d30","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"61bc7fc51d9394343d434ad8ecb8c5ac91bd544eacbf854aea228dde90169d30","first_computed_at":"2026-05-20T00:00:40.329986Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:00:40.329986Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ZN/evO+FvAW3kKnkb+CZ9GXYed/kVvCtvAfcn38IIqcErSiD7xkOgekGYqfwt875PyPYSkii20w67KNeT0qFDA==","signature_status":"signed_v1","signed_at":"2026-05-20T00:00:40.330761Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.01970","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:655e5425fa2bfa4a6bc890e1dda0557bb63b6963da12e07a57f11f76b793472e","sha256:a34f3fd04ab357a5c150efe73cdc2ddaf15e34d8ef9591f3df1f69f1353b0110"],"state_sha256":"09982b10b11c6c5a7c32b846c77a2592f5cdc9a380157926a2c66efc241dc972"}