Independent security researchers used Anthropic's Claude to access an OpenAI
employee's ChatGPT account, read and propose edits to the company's private
software cache. The researchers were participating in OpenAI’s bug‑bounty
program, reported the finding promptly and received a $6,500 bounty. The
incident follows an AI-agent breach of Hugging Face two weeks earlier and adds
to a string of recent intrusions using rapidly evolving AI tools, underscoring
defensive challenges posed by complex modern systems.