Google DeepMind describes AlphaProof Code, combining human-approved specifications with agent-generated proofs checked by Lean. Its libsignal study proves specified properties of PQXDH and the Triple Ratchet at the AArch64 machine-code level. Assurance depends on the specifications, formal semantics and other stated trust assumptions; it does not cover every property of the Signal application.
Why it matters
It supplies machine-checkable assurance beyond generated patches and tests, while making the remaining trusted components explicit.
Key facts
As stated in the sources, with where to find them.
- The framework uses human-guided specification writing and Gemini Carbon prover agents. Proofs must pass Lean’s kernel and match the approved specification.Technical paper: Figure 1 and introduction
- The libsignal proofs cover PQXDH and the Triple Ratchet, establishing forward secrecy, authentication, post-compromise security and handshake secrecy for the analyzed implementation.Technical paper: abstract and security results
- The paper reports about 6.5 million lines of machine-checked Lean proof code resting on about 26,000 lines of trusted specifications and AArch64 semantics.Technical paper: abstract
- The results additionally trust Lean’s kernel, cryptographic assumptions and contracts for some called routines. A proof establishes the stated properties under these assumptions, not blanket application security.Technical paper: Figure 1; section 4
- Reintroducing two previously fixed bugs into isolated code produced machine-checked insecurity proofs in both cases. The authors released verification artifacts for community inspection.Release: How we verified Signal; technical paper
Findings that cite this record
No tracked finding cites this record yet.
Key questions this bears on
- Can AI agents defend and oversee systems on their own?Not reliably. Some designs improve benchmark triage, but autonomous defense lacks operational validation and production oversight of agents has documented gaps.
Sources
Related records
Oct 13, 2026
Oct 12, 2026
Sep 28, 2026
Oct 27, 2026
Sep 30, 2026
Oct 21, 2026