pith. sign in

System prompt poisoning: Persistent attacks on large language models beyond user injection

2 Pith papers cite this work. Polarity classification is still indexing.

2 Pith papers citing it

citation-role summary

background 1

citation-polarity summary

fields

cs.AI 1 cs.CR 1

years

2026 1 2025 1

verdicts

UNVERDICTED 2

roles

background 1

polarities

background 1

representative citing papers

The Surface You Test Is Not the Surface That Breaks

cs.CR · 2026-05-28 · unverdicted · novelty 6.0

Prompt injection vulnerability in tool-augmented LLMs is a model-surface interaction rather than a fixed channel property; the same payload inverts success rates across models, and adaptive attack rate exceeds single-surface baselines by 9.1 pp on average.

citing papers explorer

Showing 2 of 2 citing papers.

  • The Surface You Test Is Not the Surface That Breaks cs.CR · 2026-05-28 · unverdicted · none · ref 13

    Prompt injection vulnerability in tool-augmented LLMs is a model-surface interaction rather than a fixed channel property; the same payload inverts success rates across models, and adaptive attack rate exceeds single-surface baselines by 9.1 pp on average.

  • Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges cs.AI · 2025-10-27 · unverdicted · none · ref 76

    A survey that taxonomizes threats to agentic AI, reviews benchmarks and evaluation methods, discusses technical and governance defenses, and identifies open challenges.