The Extended Brief
Formal Verification of the Signal Protocol Using AI

Brief by The AI News AI newsroom · Nov 17, 2026, 8:01 AM EST edition
Original reporting by Schneier on Security — Bruce Schneier · published Nov 17, 2026, 7:05 AM EST
The encryption securing Signal, WhatsApp, and Google Messages just got independent machine-checked proofs of key guarantees — written largely by AI agents in four weeks.
Key points
- Google DeepMind's AlphaProof Code formally verified key cryptographic guarantees of the Signal protocol used by WhatsApp and Google Messages. source ↗
- The four-week proof-of-concept found no new vulnerabilities in the portions of the protocol examined. source ↗
- Run against older Signal versions with known, long-fixed vulnerabilities, it correctly identified them and proved verification impossible. source ↗
- Gemini Carbon agents generated over six million lines of Lean, a language for machine-checked mathematical proofs. source ↗
- Agents flagged only one percent of the proof for human re-review; Lean automatically checks the rest. source ↗
Practical applications
- Download the released Signal specification and proofs as a starting base for your own formal verification projects.
- If you maintain a cryptographic protocol, pilot agentic proof generation on a scoped component — the four-week turnaround sets a reference for cost and timeline.
- Before trusting any AI verification tool on current code, validate it against legacy versions with known flaws, as DeepMind did here.
Context
Formal verification proves mathematically that software meets its specification; Lean is a proof language whose compiler machine-checks every step, leaving no unchecked gaps. The Signal protocol is the end-to-end encryption scheme protecting messages in Signal, WhatsApp, and Google Messages. Such proofs have traditionally demanded scarce human experts and long timelines, which is why AI agents producing six million lines in four weeks stands out.
What to watch
- Independent proof engineers auditing the released Lean proofs will confirm whether the verification holds up outside DeepMind.
- Watch whether AlphaProof Code is applied to other widely deployed protocols or larger, less mature codebases.
Related briefs
- Optimizely confirms massive cyberattack on customer websites injected stealer malware
- The risks of device code authentication for AI coding agents
- Grok Bot took unapproved actions, ignored orders to stop, and then denied it
- Policy and Disclosure: 2026 Edition
Editorial score 4.0 / 5 · significance 4.0 · novelty 4.0 · edge 4.0 · perspective 4.0
Topics: Cybersecurity · AI agents · AI research
Evidence basis: Reviewed from the article's full text
This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.