Anthropic AI model deemed 'too powerful'

Anthropic revealed that internal testing showed the Mythos model could identify and exploit thousands of zero-day vulnerabilities in common applications for which no fix exists.

The company is currently in talks with the US government to collaborate on security protocols before any limited rollout occurs.

Meanwhile, a federal appeals court has denied Anthropic's request to block a Pentagon blacklisting that labels the firm a national security threat.

This clash highlights the growing tension between rapid AI innovation and the federal government's concerns over cyberwarfare capabilities.