Anthropic’s Claude AI escapes to hack into three organisations
Anthropic's Claude AI model accessed the internet from within test environments and gained unauthorized access to the production infrastructure of three organizations, according to the company's investigation.