DeepMind's AlphaProof Code verifies Signal encryption at machine-code level with 6.5M lines of Lean proofs
mark_k · x · 2026-11-11
Google DeepMind introduced AlphaProof Code, a multi-agent AI framework for formal software verification, and used it to prove key security properties of Signal's libsignal implementation down to machine-code level.
- A human-guided specification agent plus autonomous prover agents generated 6.5 million lines of machine-checked Lean proofs, resting on 26,000 lines of trusted specifications and formal AArch64 semantics.
- The proofs cover Signal's post-quantum PQXDH handshake and Triple Ratchet, establishing forward secrecy, authentication, and post-compromise security — the first proofs connecting protocol-level guarantees to unmodified code of a widely used cryptographic implementation.
- The work took four weeks, a task that could have taken expert teams years; the team also reintroduced two previously fixed Signal bugs and the system proved both violated the security guarantees.
- Proofs cover specific properties under defined assumptions, but the authors argue this makes formal verification practical for real-world software. Code is open-sourced.
More from Safety
- LMArena co-founder explains AI evaluation methodology and policy in new podcast — MattPerault · 2026-11-11
- Defending METR: public data can't second-guess its AI speedup findings — peterwildeford · 2026-11-11
- Australia's review of an OpenAI agent's unauthorized Medicare access exposes the agent accountability problem — cmb24k · 2026-11-11
- Adversarial RL envs for backdoor insertion vs. detection: where does it converge? — DimitrisPapail · 2026-11-11
- Using a third-party inference provider with your coding harness hands it shell access to your machine — TheZachMueller · 2026-11-11
- Democrats propose requiring federal permission to train AI models, with up to 20 years in prison — Polymarket · 2026-11-11