A group of U.S. Senators has sent a letter to OpenAI CEO Sam Altman asking for detailed information about the company's safety practices and working conditions by August 13, 2024.
The letter is a response to media reports, some from former employees acting as whistleblowers, that highlight potential safety risks at OpenAI. The New York Times, Vox, and the Washington Post have reported on the departures of several AI safety researchers, vulnerabilities that led to a hack, and employee concerns about OpenAI's safety practices.
Some researchers, such as former super AI safety chief Jan Leike, have publicly voiced harsh criticism, stating that OpenAI prioritizes product releases over safety.
The new AI model, GPT-4o, was reportedly rushed through safety testing in just one week. It can be manipulated to generate malicious content, such as bomb-making instructions, simply by phrasing prompts in the past tense.
The Senators note that OpenAI has publicly committed to the safe and responsible development of AI and has made commitments to the Biden Administration to do so. OpenAI is also working with U.S. agencies to develop AI-based cybersecurity tools to protect critical infrastructure.
The public must be able to trust that OpenAI is developing its systems safely, the senators say. This applies to the integrity of its corporate governance, security testing, hiring practices, adherence to public promises, and cybersecurity policies.
Possibly in response to the letter, OpenAI has released several statements via X, referring to the recently established Safety and Security Committee, the leaked and now officially confirmed five levels of AGI, its Preparedness Framework, and the process of revising the much-criticized employment contracts that restricted the rights of former employees acting as whistleblowers.
