The U.S. presidential administration does not plan to publicly disclose the new evaluation system for advanced artificial intelligence models, which will determine the rules for government interaction with AI developers. This is reported by Axios, citing three sources familiar with the discussions, writes UNN.
Details
According to the publication, only companies participating in the relevant process will have access to the details of the voluntary system. As a result, other industry representatives, scientists, politicians, and U.S. allies will not know the criteria by which Washington will evaluate the most powerful artificial intelligence models.
New transparency rules for artificial intelligence systems come into force in the EU03.08.26, 15:13
On Tuesday, the White House held closed meetings with industry representatives to discuss the completed framework. At the same time, companies that did not receive an invitation have no information about its content.
The document will regulate access to advanced AI models
As Axios notes, the new system is being developed against the backdrop of an increase in large-scale cyberattacks and the rapid development of Chinese artificial intelligence models, which has reignited the debate over the safety of open AI models.
According to sources, the discussions also raised the issue of open-source models. Representatives of Nvidia participated in the meetings.
The document is expected to define the procedures for developers' cooperation with the federal government, as well as rules for providing the government with access to new artificial intelligence models 30 days before their release. In particular, it concerns requirements in the areas of cybersecurity, privacy, intellectual property protection, insider access risks, and non-disclosure of information.
Alibaba joined ranks of Chinese AI breakthroughs with new Qwen model03.08.26, 10:54