Anthropic says its Claude models 'gained unauthorized access' to other organizations' systems

Anthropic said it discovered three instances where its Claude AI models accessed the internet during an evaluation and accessed outside systems.

Kategori: Jeopolitik