P098 AI-enabled verification
When will AI auditors be competent and secure enough to deploy, and can they resist jailbreaks, prompt injection and backdoors by a sophisticated adversary?
AI-enabled verification and auditor robustness.
Status
Started A built LLM-auditor prototype (TEE and attestation emulated on a CPU without SEV; frontier model called remotely) evaluated on 21 real codebases: perfect discrimination of correct from wrong papers on objective questions, less stable on subjective ones; the audit task is software-artifact verification, not compute verification. The 'when' clause has no timeline estimate. Resistance to jailbreaks, prompt injection and backdoors: not tested; the author concedes prompt injection cannot be prevented and designs for it to leave evidence in a signed transcript.
We demonstrate this architecture by automating the artifact evaluation process for 21 peer-reviewed computer science papers with released codebases on GitHub (e.g. Does the codebase implement the system described in the paper?). We verified five high-level properties of these codebases described in the corresponding publications, treating the source code as private.
Posed by
- Baker, Kulp, Marks, Brundage & Heim (2025), Verifying International Agreements on AI: Six Layers of Verification for Rules on Large-Scale AI Development and Deployment (RAND WR-A4077-1):
However, it is unclear when AI agents will have the needed capabilities and reliability, and deployment poses logistical and cybersecurity challenges.
(§4.5, p. 33, L2007–2009)
Also raised in: Scher & Thiergart (2024), Mechanisms to Verify International Agreements About AI Development (MIRI); Hausenloy & Li (2026), The Compute Verification Post.
Sub-questions
None with a checkable form.
Sub-questions not yet stated precisely enough to check
| ID | Sub-question |
|---|---|
| P098.1 | When will AI auditors (zero-knowledge review, 'Evaluations Bot', agent-written tests) be competent, reliable and secure enough to deploy for verification? |
| P098.2 | Can AI-agent auditors resist jailbreaks, prompt injection and backdoors by a sophisticated adversary? |
Possibly relevant work found by automated search
Searched; nothing found for these sub-questions.
Data
- problems.csv — one row per problem
- problem_status.csv — status, scope and evidence per problem
- problem_sources.csv — the posed-by quote and other documents per problem
- leaves.csv — all sub-questions, including those without a checkable form
- leaf_sources.csv — every source row behind each sub-question