AI models from Gemini, Claude and ChatGPT resorted to nuclear escalation in the majority of simulations when placed in charge of nuclear-armed powers, according to the research from King’s College London (Getty/iStock)
Claude, the Anthropic AI chatbot, has gone down in a major outage.
The company said that users were seeing an “elevated rate of errors” when using Sonnet 4.6, the model that powers Claude as well as other parts of Anthropic’s offering.
In practice, that meant that the system would get stuck seeming to think, without giving any response to a question.
The problems followed issues on Tuesday, which also resulted in errors showing to users. Anthropic said then that it had fixed the issue and the service had returned to normal.
