Anthropic expands security program, opens Claude Mythos to 200 organizations

Anthropic has expanded its Project Glasswing security program, making the Claude Mythos model available to about 200 organizations across 15 countries. The model debuted in April, initially limited to a few partners like Apple.

Project Glasswing lets companies use Claude Mythos Preview to find security bugs in operating systems and web browsers, helping human researchers audit code. New participants include firms from Canada, Australia, France, Germany, Japan, South Korea, and more.

Anthropic uses a controlled release approach, requiring new partners to meet security standards before accessing Claude Mythos. Samsung and NATO may join the list, with more communication and hardware companies involved.