Anthropic's latest Claude model can interact with computers - what could go wrong?
For starters, it could launch a prompt injection attack on itself...
The latest version of AI startup Anthropic's Claude 3.5 Sonnet model can use computers ... Read more