OpenAI Rolls Out Sol Model
July 10, 12:00 PM
OpenAI has begun wide deployment of its artificial intelligence model Sol, which experts consider at least as capable as Anthropic’s Fable model. However, the process for evaluating AI safety in the United States remains opaque. Specific rules and the responsible agency for safety assessments have not been defined, and the final mechanism must be agreed upon by early August.
Issues with the Fable Model
Anthropic’s Fable model was previously temporarily restricted due to U.S. government concerns that it could be used for dangerous hacking tasks. The U.S. government barred foreign nationals from using this model. Anthropic reported cooperating with the government to create a system for detecting attempts to bypass restrictions and implementing multiple layers of protection.
Last month, the U.S. administration issued an executive order outlining a general plan for evaluating advanced AI models. Yet, as experts note, significant uncertainty surrounds the process. Mina Narayanan stated that
“we don’t have enough information to assess whether the current procedure is robust enough,”and Dean Ball also expressed concern, noting that
“nobody knows exactly what requirements need to be met to get approval for releasing such models.”Andy Konwinski added that
“even employees at leading labs don’t understand how these decisions are made.”
OpenAI CEO Sam Altman discussed the Sol launch with Commerce Secretary Howard Lutnick, Treasury Secretary Scott Bessent, and Cybersecurity Director Sean Cairncross. OpenAI also cites independent audits conducted by the UK’s AI Safety Institute (AISI), SecureBio, and Irregular. Before the official launch, the Sol model was demonstrated to government representatives and a limited group of users, but their identities have not been disclosed.
According to media reports, Altman offered to transfer up to 5% of OpenAI shares to the administration as part of the Trump Accounts project. Notably, OpenAI President Greg Brockman is the largest known donor to Trump’s political campaign. These circumstances highlight the complex interplay between technology companies and government agencies in the field of AI safety.
The launch of OpenAI’s Sol model amid a lack of transparency in U.S. AI safety evaluation underscores the need for clear rules and standards in this area. Government restrictions on models like Fable reflect growing concerns about potential risks. The interaction between tech companies and state bodies in the context of AI safety could significantly influence the future development and regulation of these technologies.
As OpenAI moves forward with the launch of its Sol model, the implications of regulatory challenges surrounding AI technologies become increasingly significant. In a related development, the Trump administration previously called for a delay in the release of new AI models, highlighting ongoing concerns about safety and oversight in the rapidly evolving landscape of artificial intelligence.