source: techcrunch ai: how did the government decide openai’s frontier model was safe to release?
level: business
openai's latest model, sol, is now publicly available, but how the government decided it was safe remains a mystery. the model is considered comparable to anthropic's fable, which was briefly banned due to security concerns. experts and even industry insiders say they lack visibility into the evaluation process. mina narayanan of georgetown's center for security and emerging technology noted that the exact dialog between companies and the government is unclear. former trump advisor dean ball wrote that nobody knows the licensing requirements.
eighteen months into the administration, there is still no formal framework. an executive order outlined a roadmap, but specifics are missing, and there will be no dedicated regulatory agency. the department of commerce is taking a temporary lead, with six cabinet agencies told to finalize a process by early august. in the meantime, openai ceo sam altman mentioned conversations with top officials, but details on expert testing remain undisclosed. openai pointed to external safety evaluations but declined to share government process details.
the ad hoc approach raises concerns about favoritism and conflicts of interest, given altman's reported equity offers and openai president greg brockman's political donations. critics argue that true technical experts are sidelined, and the secrecy undermines public trust. proposals for improvement include third-party auditing and open commons models like the fda. without transparency, the industry faces political backlash as americans grow skeptical of ai development.
why it matters: unclear safety evaluations for powerful ai models could lead to unchecked risks and erode public trust in ai governance.
source: techcrunch ai: how did the government decide openai’s frontier model was safe to release?