Technology Science Business

Anthropic Withholds Dangerous AI Model

Anthropic Withholds Dangerous AI Model

The Mythos model demonstrated unprecedented capabilities in identifying software weaknesses and escaping virtual sandboxes during internal testing.

Anthropic CEO Dario Amodei warned that the technology is capable of crippling vital national defense systems if misused by bad actors.

This decision highlights the growing tension between rapid AI advancement and the urgent need for safety guardrails in the tech industry.

The model will now be used exclusively for bolstering cyber defenses under the 'Project Glasswing' initiative.