Secret Rules for AI: A Recipe for Disaster

The White House’s decision to keep its proposed evaluation framework for frontier AI models under wraps is a stark reminder of the government’s penchant for secrecy when it comes to regulating emerging technologies. The fact that major tech companies like Meta, Nvidia, Microsoft, OpenAI, and Anthropic were invited to review the draft proposal without any public input or transparency raises more questions than answers.

The administration’s executive order mandating the creation of this framework within 60 days has already been met with skepticism from experts who warn that a voluntary process is unlikely to yield meaningful results. Chris McGuire, senior fellow for China and emerging technologies at the Council on Foreign Relations, described the decision as “baffling.” This irony is not lost, given these companies’ cooperation with government agencies reviewing their latest model releases, including Anthropic’s Mythos 5 and Fable 5 models, which were temporarily pulled from the market due to export controls.

The lack of transparency surrounding this framework may have far-reaching consequences. Frontier AI models are already raising concerns about their potential misuse, as seen in recent hacks by OpenAI and Anthropic that compromised Hugging Face’s systems. The public deserves to know how these models will be vetted and secured before they’re released into the wild.

Without a clear set of rules or penalties, it’s unclear what incentive companies have to comply with the proposed framework. This lack of transparency is particularly puzzling given the government’s recent efforts to collaborate with industry leaders on AI regulation. The administration’s decision to keep this information private fuels speculation and distrust among experts and the public.

In an era where AI is increasingly becoming a crucial component of national security, defense, and economic systems, it’s imperative that the public has access to accurate information about how these technologies are being developed and regulated. The secrecy surrounding this framework undermines the government’s credibility in this area.

The current discussions on Capitol Hill regarding formalizing engagements with industry leaders on AI regulation are a step in the right direction. However, without transparency and clear guidelines, it’s difficult to gauge the effectiveness of these efforts. A more open approach would foster public trust and encourage meaningful collaboration between government agencies and tech companies.

As the administration moves forward, prioritizing transparency and accountability when developing regulations for emerging technologies like AI is essential. The consequences of inaction or secrecy are too great to ignore. With the latest developments on Capitol Hill, it will be interesting to see if this framework is finalized and made available to the public.