Anthropic's Claude AI autonomously discovered 500+ critical vulnerabilities in popular open-source software using only basic ...

“Once contribution and reputation building can be automated, the attack surface moves from the code to the governance process around it. Projects that rely on informal trust and maintainer intuition ...

Hundreds of popular add‑ons used encrypted, URL‑sized payloads to send search queries, referrers, and timestamps to outside servers, in some cases tied to data brokers and unknown operators.