University of Cambridge

Researchers in the University of Cambridge's Department of Computer Science and Technology developed Attestable Audits, protocols for binding AI safety benchmark results and later inference to the same model S-0009:

• The Attestable Audits paper runs AI safety benchmarks inside a trusted execution environment and publishes an attestation that binds the model hash, the audit code and data, and the result to a transparency log S-0009; see Attestable Audits and Confidential multi-party verification.
• In the paper's inference protocol, each response carries an attestation that links the model, its earlier audit result, the prompt and the response S-0009; see Safeguard attestation and Model identity attestation.
• A prototype ran a 4-bit, 8-billion-parameter model in CPU-only AWS Nitro Enclaves S-0009; see TEE remote attestation for AI workloads.