US to Review Closed AI Systems before Rollout under Trump Plan
In a conference with leading artificial intelligence firms, the federal government intended to evaluate only specific kinds of artificial intelligence models for possible security threats and exclude others, as per reports.
The A.I. models the government intends to evaluate are referred to as “closed” models, which do not disclose their underlying code and are developed by firms such as Anthropic and OpenAI, based on reports.
The administration currently does not intend to evaluate “open source” A.I. models, which provide computer code for public access and modification; however, this could shift as the technology evolves, according to reports.
The actions represent an attempt by the Trump administration to adopt a more active stance on A.I. regulation, establishing formal oversight of major A.I. laboratories such as OpenAI and Anthropic. However, they postpone tackling certain urgent regulatory concerns related to the technology, such as the potential risks posed by more potent open-source models developed by Chinese firms.
The choice to avoid open-source models may benefit companies aiming to surpass Anthropic and OpenAI, regarded as frontrunners in next-gen A.I. development
The structure, emerging after extensive confidential talks among government representatives and a select group of the world's most influential technology firms, is poised to influence the global A.I. competition. For a long time, companies have predominantly made their own decisions about granting the public access to recent advancements or keeping them secret due to safety and security concerns.
Also Read: AstraZeneca, Bristol Myers in $400 billion Pharma Merger Talks
At this point, the United States will have an enhanced chance to influence the situation, possibly allowing the government to delay the implementation of the technology.
Present at the meeting were delegates from Anthropic, OpenAI, Microsoft, Meta, Google, and Nvidia, as noted by two attendees. The White House requested input from the companies, but the regulations for the A.I. evaluations are nearly finished, according to reports.
Also Read: Amazon Hits $3 Trillion Market Cap on AI-Driven Stock Surge
Nvidia, Meta, and Google, among other American firms, develop open-source models. However, it is the open-source models from Chinese companies like Alibaba, DeepSeek, and Moonshot AI that have raised the greatest alarm among administration officials.
Also Read: Unitree Technology Eyes $7.4 Billion Valuation Ahead of Shanghai IPO
Chinese models have recently achieved results close to those of top American models, raising fears that China might be advancing in the A.I. competition. However, it is uncertain if the United States can effectively regulate this foreign technology via the new review system, and it is improbable that Chinese firms or Beijing would comply with American evaluations.