Anthropic Hides Powerful Cyber AI From Hackers

Anthropic unveiled Claude Mythos, a cybersecurity AI model so capable at finding vulnerabilities that the company is restricting its release to a select group called Project Glasswing. The move aims to help developers build more secure software before bad actors can exploit the technology, though skepticism remains about keeping it contained. Meanwhile, enterprises are pushing back against AI chaos, demanding more control over agents and deployments. Meta re-entered the AI race with its Muse Spark reasoning model, while analysts warn that AI's compute-heavy, capital-intensive nature means the current boom may not prove profitable for all players involved.