P121 Building trust in shared mechanism designs
Does shared design documentation plus adversarial review build enough international trust in shared mechanisms, or will states insist on de novo designs?
Building international trust in shared mechanism designs and standards.
Status
Analyzed A qualitative desk comparison of three routes to a trusted standard (adapting existing commercial designs, open-source projects such as OpenTitan, seL4 and Caliptra, and a de novo international collaboration), weighing the security limits of commercial designs and the precedent of weakened encryption standards; no review conducted, no standard written, no state's position recorded. Standards for assessing other parties' solutions, trust in lab findings and an appliance standard: calls for work only.
"Concerns about subtle vulnerabilities could likely be addressed through a detailed international review of the design, where each design choice is carefully justified to ensure that it was not animated by any ulterior motive. However, sufficiently suspicious states may still insist that the only acceptable long-term solution is a de novo design."
Posed by
- Petrie, Aarne, Ammann & Dalrymple (2025), Overview of Flexible Hardware-Enabled Guarantees (flexHEG Part 1):
There would need to be an effort to establish international trust in the current design, by sharing the design documents and source code, and reviewing each design decision in detail with experts on the other side.
(Transition to Intl Governance, p13, L532–534)
Also raised in: Petrie & Aarne (2025), Technical Options for Flexible Hardware-Enabled Guarantees (flexHEG Part 2); Aarne & Petrie (2025), International Security Applications of Flexible Hardware-Enabled Guarantees (flexHEG Part 3); Cankaya (2026), Bit-Exact AI Inference Verification Without Performance Tradeoffs, arXiv 2606.00279v2; Abdelghafar & Kulp (2026), Privacy-Preserving AI Verification via Minimal Information Disclosure; Karvonen et al. (2025), DiFR: Inference Verification Despite Nondeterminism; AI Security Forum, Requests for Discussion; Dean, Verification Plan (AI 2040); Proofworks, Living Verification Documents.
Sub-questions
| ID | Sub-question | What the sources report |
|---|---|---|
| P121.1 | Can a Groth16 circuit-specific trusted setup and verifiable sampling for randomized releases be run acceptably between parties? | partly answered |
| P121.2 | Should providers be required to expose per-request seeds with a documented sampler, and how many of the top inference APIs do? | open |
| P121.3 | What throughput cost does a mandated batch-size-sensitive (non-invariant) kernel set impose vs best invariant kernels? | open |
Sub-questions not yet stated precisely enough to check
| ID | Sub-question |
|---|---|
| P121.4 | Does shared design documentation plus adversarial review build sufficient international trust, or will suspicious states insist on a de novo design? |
| P121.5 | What standards can credibly assess the trustworthiness of verification solutions supplied by other parties? |
| P121.6 | How can international trust in hardware-lab findings be established? |
| P121.7 | Is there a standard defining what an AI appliance is? |
| P121.8 | Who must share networking/compute paradigms so compatible taps and recomputation servers can be built? |
| P121.9 | Who declares the authorised target, protected properties and utility threshold before mechanism selection? |
Possibly relevant work found by automated search
Not searched: no technical result was expected for this problem.
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