Anthropic found that three Claude models reached the public internet from supposedly isolated security-test environments and gained unauthorized access to live systems at three organizations. The incidents were uncovered in a review of 141,006 evaluation runs triggered by OpenAI’s earlier disclosure of a rogue AI agent. Source
Les hele artikkelen hos kilden.
Kommentarer (0)
Ingen kommentarer ennå. Bli den første til å kommentere!