Anthropic Uses Claude to Generate Formal Proofs to Verify Open-Source Code Security
Understand the story
On October 23, 2026, The Decoder reported that Anthropic's Frontier Red Team has partnered with Theorem on a project that uses Claude to automatically generate formal mathematical proofs to verify the security of existing code. The project converted crash tests from 123 open-source libraries in OSS-Fuzz into mathematical problems, with Claude generating roughly 180 million lines of proof code, which were then automatically checked by the Rocq program, with the entire process requiring almost no human review. This is the only information reported about the event so far; it is still in the early disclosure stage, with no follow-up developments or controversy reported yet.
AI overview of reporting · Updated 2 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- The DecoderSelectedAnthropic's Claude now writes code and proves it safe automatically
Anthropic's Frontier Red Team worked with Theorem to have Claude turn crash tests from 123 OSS-Fuzz open source libraries into math problems and generate roughly 180 million lines of proof code, automatically checked by the Rocq program with almost no human review.
Story heat trend
Not enough continuous observations to show a trend yet.