AI treaty verification may need three layers: pragmatic checks, enclaves, and math
geoffreyirving · x · 2026-07-25
Geoffrey Irving argues that the near future may require AI treaty verification technology, and outlines three rough tiers of approaches.
- Pragmatic: operational methods that avoid looking at the code directly, such as datacenter inspections, hardware tracking, and chip controls.
- Enclaves: run agreed code inside secure enclaves; he notes some accelerators, including Nvidia GPUs, already have enclave support and can keep slowdown low.
- Math: cryptographic circuit obfuscation, combining FHE for nonlinear parts with custom techniques for matmuls; he says today’s slowdown is still several orders of magnitude too high.
He adds that hardware, security, and crypto researchers should work across all three levels, since each has different trade-offs against side channels and physical attacks.
Related event: Geoffrey Irving Outlines Three-Layer Approach to AI Treaty Verification(7 posts)→
More from Safety
- rauchg reproduces OpenAI's covert CPU-channel finding: chat across Firecracker microVMs with no network — ccerrato147 · 2026-10-27
- Dev questions reports of Chinese models writing insecure code, citing thin methods and conflicts — dbreunig · 2026-10-27
- Aidan Gomez: LLM probing outside servers was out-of-bounds, not proof of intent — AccBalanced · 2026-10-27
- Why a patched PC still loaded the BlackX UEFI bootkit — and how to check yours — titusblair · 2026-10-27
- Exponential launches RepoInject and BrowserInject benchmarks for prompt injection robustness — maksym_andr · 2026-10-27
- Poster surprised that v4.1 flash gets more accurate when asked to control its CoT — dfrsrchtwts · 2026-10-27