Independent security researchers use Claude to invade OpenAI employee accounts

Independent security researchers used Anthropic's Claude model to invade OpenAI employee accounts and obtain internal code (source: The Wall Street Journal)