Anthropic warning on AI: close to self-building capability
In a revised US securities filing, Anthropic said it may soon meet its own bar for dramatically accelerated AI progress, while pledging that humans will stay in charge of frontier model work.
Anthropic’s revised S-1A risk factors say the company may soon meet its own definition of dramatically accelerated progress driven by AI.
The firm has already used AI to help build newer models and says some capability jumps were not spotted until after they occurred.
No current Anthropic model has been marked as crossing that internal threshold, the filing states.
Anthropic says using AI to build more capable AI can raise safety risks if progress speeds up sharply.
It is weighing Responsible Scaling Policy updates so humans remain in control of frontier model development.