Sam Altman revealed in a tweet that OpenAI will provide early access to its next model to the US AI Safety Institute as part of its safety efforts. The company has been collaborating with the consortium to advance the science of AI evaluations. The National Institute of Standards and Technology (NIST) officially established the Artificial Intelligence Safety Institute earlier this year, after Vice President Kamala Harris announced it back in 2023 at the UK AI Safety Summit. According to NIST’s description, the consortium aims to develop science-based guidelines and standards for AI measurement and policy to enhance AI safety globally. OpenAI, along with DeepMind, also pledged to share AI models with the UK government last year. Concerns have arisen regarding OpenAI’s safety priorities as it focuses on developing more powerful AI models, leading to speculation about leadership changes within the company. Despite these concerns, OpenAI has disbanded the Superalignment team and established a new safety group led by board members like Altman, raising questions about self-policing within the organization.
