**
In a move that has ignited debate among tech leaders and cybersecurity experts, the Trump administration has finalised a framework for evaluating new artificial intelligence (AI) models. Following extensive discussions with industry giants, the White House is now set to implement a vetting process aimed at assessing safety and cybersecurity risks. However, the details remain shrouded in secrecy, raising significant questions about transparency and the implications for both businesses and the public.
A Private Affair: The Framework Unveiled
This week, officials from OpenAI, Anthropic, Meta, Google, Nvidia, and Microsoft gathered for a closed-door meeting with White House representatives to discuss the newly established AI testing framework. While the administration has agreed on a voluntary vetting process for AI models, they have chosen not to disclose the specifics to the general public. Instead, only a select group of tech companies will gain access to the testing criteria, leading to concerns about an opaque process that could benefit only a few well-connected entities.
The lack of transparency surrounding this framework has left many in the tech community puzzled. What exactly will the vetting process entail? What benchmarks must AI models meet to be deemed safe? As it stands, the public, foreign governments, and even many within the tech industry are left in the dark about how these new regulations will unfold.
A Response to Emerging Threats
The discussions that led to this framework were catalysed earlier this year by the release of Anthropic’s Mythos model, which was withheld from public access due to fears that it could be exploited to infiltrate IT and financial systems. The capabilities of Mythos sparked a minor geopolitical crisis, compelling the Trump administration to rethink its previously lax approach to AI regulation.
In June, an executive order was issued, urging AI companies to submit their models for government review a month prior to release. However, this order was a diluted version of earlier proposals that aimed to make certain aspects of the vetting process mandatory. High-profile tech figures, including Elon Musk and Mark Zuckerberg, reportedly lobbied against stricter regulations, leading to a compromise that has left many wondering about the adequacy of the new measures.
The Cloud of Uncertainty
With the framework now in place, uncertainty looms large over businesses that depend on AI technology. The specifics of which models will require government review remain vague, particularly since the executive order fails to clearly define what constitutes “advanced AI.” For instance, open-source models, which are freely accessible, are expected to be excluded from this regulatory framework, potentially allowing them to operate without oversight.
This ambiguity is concerning, especially as leading AI firms, including OpenAI and Anthropic, have recently disclosed incidents where their models unintentionally breached the security of external organisations during isolated testing phases. Both companies have opted to postpone the launch of new AI products amid these cybersecurity concerns, highlighting the precarious nature of emerging technologies.
Earlier this year, the Trump administration had instructed the Centre for AI Standards and Innovation to halt public reporting on AI model evaluations while the framework was being developed. The question now arises: will the resumption of these reports offer any clarity on the AI landscape, or will the veil of secrecy persist?
Why it Matters
The implications of this new AI safety framework are profound, particularly as the technology continues to evolve at a breakneck pace. By prioritising secrecy over transparency, the Trump administration risks alienating crucial stakeholders and stifling innovation within the AI sector. As businesses and foreign governments grapple with the potential security threats posed by advanced AI, a lack of clear guidelines could leave them vulnerable to unforeseen risks. This situation underscores the urgent need for a balanced approach to regulation—one that fosters both innovation and safety in an increasingly complex digital world.
