OpenAI has commenced the public rollout of its latest advanced language model, Sol, which is considered to be at least on par with Anthropic’s Fable. The release of such powerful AI models has raised questions about the government’s approval process, as the specifics remain ambiguous.
Experts in the field have expressed concerns over the lack of transparency. Mina Narayanan, a senior research analyst at Georgetown’s Center for Security and Emerging Technology, stated that she lacks visibility into the exact processes and cannot assess their adequacy. Similarly, Dean W. Ball, a former policy advisor now at OpenAI, noted that the requirements for obtaining a license are unclear.
Andy Konwinski, co-founder of Databricks and Perplexity, highlighted the existential problem of decision-making power, emphasizing the need for clarity on who has the authority to grant permissions for such releases.
Despite the Trump administration’s tenure, there is still little clarity on the path forward. An executive order was recently published, outlining a roadmap for evaluating frontier models, but specifics are yet to be determined. Notably, Sriram Krishnan, a former senior advisor for AI in the White House, mentioned that there will not be an FDA equivalent for AI.
Currently, the Department of Commerce’s Center for AI Standards and Innovation appears to be leading the efforts, but the executive order instructs six cabinet agencies to finalize the process by early August. In the interim, the approach seems to be ad hoc.
OpenAI CEO Sam Altman mentioned that the process involved discussions with officials like Secretary of Commerce Howard Lutnick, Secretary of the Treasury Scott Bessent, and US national cyber director Sean Cairncross. However, details about the experts who tested the models and their methodologies remain undisclosed. OpenAI has pointed to external evaluations by organizations such as UK AISI, SecureBio, and Irregular in the model’s safety card but declined to share specifics on the government’s process.
As with Anthropic’s Fable rollout, OpenAI previewed the model for the government and select users ahead of the wider release. This approach underscores the ongoing challenges in establishing a clear and transparent approval process for advanced AI models.
The lack of a standardized and transparent approval process for releasing advanced AI models like Sol raises significant concerns. As AI technology continues to evolve rapidly, it is imperative for both the government and AI developers to establish clear guidelines and processes to ensure public safety and trust. Without such measures, the deployment of powerful AI systems may outpace the development of necessary regulatory frameworks, potentially leading to unforeseen risks and challenges.
