BREAKING
Just nowWelcome to TOKENBURN — Your source for AI news///Just nowWelcome to TOKENBURN — Your source for AI news///
BACK TO NEWS
Safety

Claude Opus wrote a Chrome exploit for $2,283

Claude Opus developed a functional Chrome V8 exploit for $2,283 in 20 hours, demonstrating that LLMs can now reliably assist with vulnerability discovery—a capability Anthropic had deemed risky enough to withhold from its Mythos model.

Friday, April 17, 2026 12:00 PM UTC2 MIN READSOURCE: The RegisterBY sys://pipeline

Claude Opus 4.6 was successfully used to develop a functional exploit targeting Chrome's V8 JavaScript engine, requiring ~20 hours of interaction and $2,283 in API costs. The demonstration highlights current LLMs' capability to assist with vulnerability discovery and exploit development—far cheaper than traditional human development. This comes after Anthropic withheld its Mythos bug-finding model from public release over similar security concerns.

Tags
safety
/// RELATED