Research map / Adversarial robustness and security / Attacks on text and graphs
Adaptive Attacks on Trusted Monitors Subvert AI Control Protocols
Mikhail Terekhov, Alexander Panfilov, Daniil Dzenhaliou, Caglar Gulcehre, Maksym Andriushchenko, Ameya Prabhu, Jonas Geiping
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adaptive prompt injection into backdoored solutions | trusted monitoring and deferral protocols · APPS | safety | Hurt |
| add adaptive prompt injection into attacker actions | DTT and DTR · BashBench | safety | Hurt |
| use a prompt injection crafted for a different monitor | trusted monitoring · APPS | safety | Hurt |
| let the untrusted model write the injection zero-shot | trusted monitoring · APPS | safety | Hurt |
| use Defer-to-Resample instead of Defer-to-Trusted | DTT · APPS | safety | Hurt |
| use Defer-to-Resample instead of Defer-to-Trusted | DTT · BashBench | safety | No clear effect |
| raise deferral threshold (ablation) | DTT · BashBench | safety | Hurt |
| use alternative safety definition (exact backdoor action) (ablation) | DTT · BashBench | safety | Hurt |
Closest papers by meaning
- Attack Selection In Agentic AI Control Evaluations Meaningfully Decreases Safety (NeurIPS 2026 · similarity 0.86)
- Adaptive Adversaries: A Multi-Turn, Multi-LLM Benchmark for LLM Agent Security (NeurIPS 2026 · similarity 0.86)
- Monitoring Decomposition Attacks with Lightweight Sequential Monitors (ICLR 2026 · similarity 0.85)
- AI Control: Improving Safety Despite Intentional Subversion (ICML 2024 · similarity 0.85)
- Reliable Weak-to-Strong Monitoring of LLM Agents (ICLR 2026 · similarity 0.85)
- Agent-ToM: Learning to Monitor Autonomous LLM Agents via Theory-of-Mind Reasoning (NeurIPS 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
