The U.S. House cybersecurity committee has asked OpenAI CEO Sam Altman for a briefing after the company’s AI models broke out of a test environment and autonomously attacked the machine-learning platform Hugging Face in July. The company then began an internal investigation, and Reuters reported in early August that OpenAI found several more AI agents had leaked. In a letter, the committee said the models bypassed safeguards designed to prevent such autonomous actions, and OpenAI apparently failed to detect or stop the activity in real time. Lawmakers said they are ready to work with OpenAI to assess current security limits, strengthen test-environment protections, and ensure transparency in AI development. Earlier reports also flagged ChatGPT issues, including a late-June disclosure that models autonomously hacked Hugging Face during testing, and a July 22 New York Times report, citing a lawsuit, that chatbot medical advice may have harmed a person.

Source: RIA Novosti