Source · Tier B · Preprint
Attestable Audits: Verifiable AI Safety Benchmarks Using Trusted Execution Environments
C. Schnabl, D. Hugenroth, B. Marino, A. R. Beresford. 2025. ICML 2025 Workshop on Technical AI Governance.
| Link | https://arxiv.org/abs/2506.23706 |
|---|---|
| DOI | 10.48550/arXiv.2506.23706 |
| arXiv | 2506.23706 |
| Version | arXiv v1 (2025-06-30) read on 2026-09-23. The arXiv comment reads "ICML 2024 Workshop TAIG", and the PDF and HTML carry the ICML 2025 template header ("Proceedings of the 42nd International Conference on Machine Learning ... PMLR 267"). The ICML 2025 virtual site (https://icml.cc/virtual/2025/48334) lists it as a workshop poster at the Workshop on Technical AI Governance, and it was not found in the PMLR 267 volume, so it is recorded as a workshop paper, not a main-conference publication. |
| Accessed | 2026-09-23 |
| Organization | University of Cambridge |
| Imported from | hodgkins-ai-verification-papers@c71e59ff0e8e |
| Note | Listed under "Trusted execution and attestation" in the Hodgkins bibliography (CC BY 4.0). |
Cited by
- R2Confidential multi-party verification
- R2Model identity attestation⚠
- R2Safeguard attestation
- R2TEE remote attestation for AI workloads⚠
- R2Attestable Audits
- The declared model is the one being served
- Declared safeguards were applied during inference
- Trusted execution environment (TEE)
- University of Cambridge