Google DeepMind has introduced AlphaProof Code, a system of AI agents that writes formal proofs in Lean, a language used to mathematically check that software behaves as specified. According to the company's announcement, the agents produced roughly three million lines of Lean proofs verifying the core encryption of Signal, the end-to-end encrypted messaging app. DeepMind presents the work as evidence that AI can tackle verification of complex, real-world software in depth, a task that has traditionally demanded large amounts of specialist engineering time. Formal verification offers stronger guarantees than testing. A machine-checked proof covers every input a program can receive, while tests only check the cases someone thought to write. Its cost has historically confined it to high-stakes components such as cryptographic libraries and safety-critical systems. By having agents write and check the proofs, DeepMind is positioning the approach as practical at a much larger scale. The show notes do not describe the agent architecture, the timeline, or how the proofs were matched to Signal's codebase, so the full DeepMind post is the place to confirm those details. For Australian businesses and professionals, this changes little immediately. The show notes mention no pricing or availability for AlphaProof Code, so there is nothing to buy or trial yet. Teams building security-sensitive software may find the approach worth watching as a way to reduce risk in cryptographic or data-handling code. Under the Privacy Act 1988, Australian Privacy Principle 11 requires entities to take reasonable steps to protect personal information from misuse, interference and loss. Formal verification is one possible measure, but it is not required, and this announcement creates no new obligation.