P105 Model identity across evaluation and deployment

Can a verifier confirm the evaluated, deployed and served models are the same (e.g. via attested weight/code hash) in a real pilot?

Showing evaluated, deployed and served models are the same.

Status

Started Served-model identity by recomputation on one 8xH100 server with prover and verifier on the same box; open-weight 7B-120B models; developer-run attacks (model substitution, LoRA append, append-on-trigger); 37,500 prompt runs; a fixed threshold gave honest-prover failures and short-output false negatives. The evaluated-versus-deployed clause (a TEE-signed hash of weights and code) is a design only, which Scher & Thiergart call 'mostly solved in principle'; no real pilot, which the headline asks for; handling of legitimate model versions: no work.

Our results add further evidence that Token DiFR can identify different incorrect-model attacks reliably across a small range of prompts.
Amodo (2026), An Inference Verification Prototype — Stage 1 (Amodo Design note) (2026-06-29), L235

Posed by

Also raised in: Harack et al. (2025), Verification for International AI Governance (Oxford AIGI); Aarne & Petrie (2025), International Security Applications of Flexible Hardware-Enabled Guarantees (flexHEG Part 3); Shavit (2023), What Does It Take to Catch a Chinchilla? Verifying Rules on Large-Scale Neural Network Training via Compute Monitoring; Harack (2026), AI Verification: Infrastructure for Prosperity, Governance, and Peace (Lawfare); Hausenloy & Li (2026), The Compute Verification Post; AI Security Forum, Requests for Discussion; Proofworks, Living Verification Documents.

Sub-questions

IDSub-questionWhat the sources report
P105.1Has a TEE-signed weight-and-code hash shown that an evaluated model and a deployed model are identical in a real pilot?partly answered
P105.2Can a client or verifier confirm, from API outputs or per-response attestation, that served weights match a committed hash (to a stated detection AUC) or a reported run?open
P105.3Has retroactive verification of inference-time weight snapshots against a declared hash been shown on a production stack?a source suggests an answer (not checked)

Sub-questions not yet stated precisely enough to check

IDSub-question
P105.4How should legitimate model updates and versions be handled under identity checks?

Possibly relevant work found by automated search

Found by AI agents searching for this problem's sub-questions, mostly by keyword and citation in the OpenAlex scholarly database and on the web; a second AI pass reviewed each one, and we spot-checked some by hand. They seemed reasonably related, as far as one can tell from outside each paper's field. Some come from neighbouring fields, such as arms control and financial audit, and bear on the problem by analogy. Not reviewed against the status above; a paper listed here does not change that status. 928 of 1052 quotes checked word for word against the paper's text.

For P105.1: Has a TEE-signed weight-and-code hash shown that an evaluated model and a deployed model are identical in a real pilot?

For P105.2: Can a client or verifier confirm, from API outputs or per-response attestation, that served weights match a committed hash (to a stated detection AUC) or a reported run?

For P105.3: Has retroactive verification of inference-time weight snapshots against a declared hash been shown on a production stack?

Data