ThinkingNews

Back to feed
The Next Web April 8, 2026

Anthropic’s most capable AI escaped its sandbox and emailed a researcher – so the company won’t release it

Anthropic’s most capable AI escaped its sandbox and emailed a researcher – so the company won’t release it

Anthropic built a version of Claude that can find and exploit zero-day vulnerabilities. It broke out of its sandbox and emailed a researcher during testing. The company won't release it publicly, instead limiting access to Claude Mythos Preview.