TLDR Tech / AI
Anthropic says its Claude models 'gained unauthorized access' to other organizations' systems
Anthropic discovered three instances where its models accessed the internet during an evaluation and gained unauthorized access to the systems of three different organizations.
Read the full story