Anthropic Expands Claude Mythos Preview Access to 150 Partners Across 15 Countries
The expansion signals Anthropic's expectation that cheap, fast AI with powerful cyber capabilities will become widely available within a year, and that defenders need access to cutting-edge tools before that happens.
Reporting from 1 sources: GIGAZINE.
Anthropic is expanding its Project Glasswing early access program for the Claude Mythos Preview vulnerability discovery tool from about 50 partners to roughly 150, adding organizations in power, water, healthcare, telecom, and hardware sectors across more than 15 countries. The program has already identified over 10,000 critical security flaws since its April 2026 launch.
Anthropic announced Wednesday that it is expanding Project Glasswing, the early access program for its Claude Mythos Preview vulnerability discovery model, from about 50 initial partners to roughly 150. The new partners span more than 15 countries and include industries not well covered in the first wave: power, water, healthcare, telecommunications, and hardware. Many are vendors or nonprofits that maintain codebases relied on by governments and organizations worldwide.
Since launching in April 2026, the program has found over 10,000 critical security flaws. Anthropic said a successful attack on most of these partners' codebases could affect over 100 million people and have significant implications for global and national security. The company expects that within 6 to 12 months, other AI firms will have models on par with Claude Mythos Preview and may release them without safeguards, making early adaptation by defenders urgent.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.