The latest
Anthropic is warning that the global race to build more powerful AI may need to slow down, saying its own systems are increasingly able to assist in creating the models that could replace them.
Details
• Anthropic said Claude now writes roughly 80% of the company’s code.
• The company said the model also helps propose research directions and solve open-ended technical problems.
• AI safety researchers have long warned about “recursive self-improvement” — the risk that AI systems could help build stronger successors, accelerating capabilities faster than humans can safely manage.
• Amodei said ensuring powerful AI systems are safe could take time.
• He argued it would be useful for the world to have the option to slow or temporarily pause frontier AI development if safety work falls behind.
What to watch
The warning is likely to intensify the debate between AI labs, governments and safety advocates over whether voluntary safeguards are enough — or whether frontier AI development needs harder limits before the next major leap in capability.
