OpenAI vows to provide the US government early access to its next AI model

OpenAI will give the US AI Security Institute early entry to its subsequent mannequin as a part of its security efforts, Sam Altman has revealed in a tweet. Apparently, the corporate has been working with the consortium “to push ahead the science of AI evaluations.” The Nationwide Institute of Requirements and Know-how (NIST) has formally established the Synthetic Intelligence Security Institute earlier this yr, although Vice President Kamala Harris announced it again in 2023 on the UK AI Security Summit. Primarily based on the NIST’s description of the consortium, it is meant “to develop science-based and empirically backed tips and requirements for AI measurement and coverage, laying the muse for AI security the world over.”

The corporate, together with DeepMind, equally pledged to share AI fashions with the UK government final yr. As TechCrunch notes, there have been rising considerations that OpenAI is making security much less of a precedence because it seeks to develop extra highly effective AI fashions. There have been speculations that the board determined to kick Sam Altman out of the corporate — he was in a short time reinstated — on account of security and safety considerations. Nonetheless, the corporate told staff members in an inner memo again then, that it was due to “a breakdown in communication.”

In Might this yr, OpenAI admitted that it disbanded the Superalignment team it created to make sure that humanity stays secure as the corporate advances its work on generative synthetic intelligence. Earlier than that, OpenAI co-founder and Chief Scientist Ilya Sutskever, who was one of many workforce’s leaders, left the company. Jan Leike, who was additionally one of many workforce’s leaders, give up, as properly. He stated in a collection of tweets that he had been disagreeing with OpenAI’s management concerning the core priorities of the corporate for fairly a while and that “security tradition and processes have taken a backseat to shiny merchandise.” OpenAI created a new safety group by the tip of Might, nevertheless it’s led by board members that embody Altman, prompting considerations about self-policing.