Anthropic says it nears self-building AI that raises safety concerns
Anthropic, a major player in AI research, says it's close to creating a "self-building AI" (basically, an AI that can write and upgrade its own code with barely any human help).
This could be a huge leap for science and technology, but it also brings up some big questions about how we keep things safe and under control.
Anthropic's Claude wrote over 80% code
Claude, Anthropic's AI system, has already written over 80% of the code used in the company's own systems as of May 2026.
That kind of rapid progress means safety checks need to keep up.
Anthropic says it's sticking to strict rules for responsible growth and making sure humans stay in charge.
Anthropic follows a Responsible Scaling Policy and says meaningful human supervision will remain absolutely necessary for the foreseeable future.