How did the government decide OpenAI's Sol model was safe to release?

OpenAI is releasing its latest advanced language model, Sol, for wide public access.

The release of Sol has prompted questions regarding the government's role in assessing the safety of AI models. Previous concerns about Anthropic’s Fable model, whose capabilities prompted the White House to briefly ban its public release, underscore the need for clarity.

Mina Narayanan, a senior research analyst at Georgetown’s Center for Security and Open Technology, told TechCrunch that she lacks sufficient insight into the exact processes to deem them adequate. "Anthropic said they were in conversations with the government and developed a classifier to detect jailbreak attempts, and implemented defensive gap strategies to prevent future jailbreaks," Narayanan stated, adding that the precise dialogue between the government, Anthropic, and OpenAI remains unclear.

Dean W. Ball, a former policy advisor in the Trump administration now working for OpenAI, wrote in his newsletter last month that "nobody knows what the requirements are to get licensed." Information on the basis of the government's approval for models like Sol's release remains opaque.

OpenAI's advanced language models, such as Sol, offer significant new capabilities, but their rapid development and broad availability necessitate ongoing dialogue and the establishment of clear safety standards between regulatory bodies and technology companies.