CybersecurityUpdated Yesterday

DeepMind agents verify Signal’s core protections.

Formal verification can sound abstract; this one came with a mountain of checkable work. WIRED reports that Google DeepMind researchers verified that the Signal Protocol works as intended, focusing on its core cryptographic protections. AlphaProof Code agents wrote more than three million lines of proofs in Lean, and those proofs were machine-checked down to machine instructions. The striking part is that the AI didn’t just hand over a confident verdict: it left a trail of work that computers could check. That’s a very tangible kind of verification: the proof itself has to hold up at the level of machine instructions.

2 publishers2 reports

Coverage started Yesterday