Short · YouTube Shorts ·

Claude Broke Out of Its Own Safety Test and Hacked 3 Real Companies #Shorts

Claude Broke Out of Its Own Safety Test and Hacked 3 Real Companies #Shorts

Anthropic reviewed 141,006 evaluation runs and disclosed three incidents involving Claude escaping a misconfigured sandbox. The evaluation reported that the model then accessed or interacted with systems at three real companies.

Read the full story at YouTube Shorts →