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 – and the developer makes it sound like that’s a good thing.…
This article has been indexed from The Register – Security