However, it is still unclear when – or if at all – the White House will announce the details of this process or how it will be implemented and enforced.
Read more «Nowhere else did the wind cause damage»: one broken tree – 5 damaged cars
According to the portal “Axios,” the evaluation process will apply only to closed models that are strictly controlled by their creators. The main creators of closed models are OpenAI, Anthropic, and Google.
Meanwhile, Meta and Nvidia are developing “open” models that users can download and update directly. Open models are developed by competitors in China, including DeepSeek and Moonshot, as well as Mistral in France.
It is reported that the Tuesday meeting at the White House was attended by executives from OpenAI, Anthropic, Google, Nvidia, Microsoft, and Meta.
The decision to exempt open models from evaluation is likely aimed at helping the US remain competitive against China.
However, according to the analytical center “Special Competitive Studies Project,” founded and chaired by former Google CEO Eric Schmidt, and technology vice president Martijn Rasser, there is a “deeper problem.”
“The US still lacks a legislatively established, predictable process for assessing the safety of advanced AI models,” he said, adding that the voluntary system applied only to closed models “concentrates” uncertainty on just a few companies.
“Unruly” models
The rapidly advancing AI in the United States has prompted the realization that rules need to be urgently developed to reduce and prevent risks associated with this technology. However, the Trump administration took a light, deregulatory approach in many industries, including the technology sector.
Read more Another new recreational area will soon open – the coastal park
In June, Trump signed an executive order requiring major AI developers to submit new models for administration review 30 days before their public release. It also set a 60-day deadline for the federal government to finalize the proposed system.
This deadline expired on August 1, but it was not publicly announced.
Recent cyberattacks independently carried out by OpenAI and Anthropic AI models have raised concerns about the capabilities of advanced AI models.
On July 21, OpenAI confirmed that its AI models escaped the testing environment and attacked another company – Hugging Face. About a week later, it reported that these models also attacked three more companies.
Later, on July 30, Anthropic revealed that it also identified three cases where test AI models “gained unauthorized access” to unspecified organizations.
Trump called for a balanced approach that would allow managing AI-related security risks while maintaining US competitiveness.
“We have to be careful on both fronts. We don’t want to restrict them so much that we suddenly end up in second place behind China,” Trump told reporters last week.
In recent months, DeepSeek and Moonshot have developed their powerful new AI models, which have again raised concerns about US competitiveness in the AI field.
Read more Tapinas, who revealed the earnings of «Žalgiris,» explained who will be allowed a budget overrun
