[Into the World of AI] U.S. Administration’s Influence Forces AI Big Tech to Change Model Direction

OpenAI Shifts to Phased AI Model Releases
Anthropic Hires Public Sector Executive

Major global artificial intelligence (AI) companies, including OpenAI and Anthropic, appear to be coming under the influence of the U.S. government. Before launching new AI models, these companies are now required to undergo pre-approval from the U.S. administration, and their models have even been subjected to export control measures. The industry is closely monitoring the U.S. government's move to treat AI models as a type of strategic asset.

According to the IT industry on July 11, OpenAI released its latest model, GPT-5.6, to the public on July 9 (local time). This model is officially available in three sub-models: Sol, Terra, and Luna. It was initially pre-released to select institutions on June 26. Among these, the top-tier Sol model is reported to deliver performance on par with Anthropic's cutting-edge model, Claude Mythos 5.

The reason OpenAI, which usually launches its models simultaneously, opted for a phased rollout this time was due to the U.S. administration’s executive order mandating pre-approval for advanced AI models. U.S. President Donald Trump signed an executive order on June 2, requiring AI companies to obtain government approval before launching new models. This measure was implemented because the performance of recently released AI models has increased significantly, raising concerns that they could be misused for cyberattacks and other malicious activities.

OpenAI explained that changes were made to the model during the U.S. government's review process. CEO Sam Altman, in an interview with local media, stated that he coordinated opinions with Secretary of Commerce Howard Lutnick, Treasury Secretary Scott Bessent, and National Cyber Director Sean Cairncross, noting that "many changes were incorporated."

CEO Altman also indicated acceptance of the government’s pre-approval process. He said, "I think it's fine as long as the process is easy to understand, fair, and quick," adding, "When we develop the next model, things will proceed much more smoothly." Previously, following the U.S. administration’s move to mandate AI model approval, OpenAI had expressed the opinion that "government approval procedures should not become a long-term standard," but this latest statement marks a step back from that position.

As the U.S. administration exerts greater influence over its domestic AI companies, its impact is growing. According to major foreign media outlets such as the Financial Times (FT), CEO Altman reportedly proposed offering a 5% stake to the U.S. administration during initial discussions, aiming to foster a cooperative relationship. However, CEO Altman dismissed these reports, saying, "Much of the information is inaccurate."

Anthropic has also faced difficulties in launching and operating its models due to conflicts with the U.S. government. Anthropic’s Claude was the only AI model that could be used for classified work by the U.S. Department of Defense (War Department), but after disagreements over the scope of the model’s use, it was designated as a supply chain risk entity and removed. At the time, the U.S. government wanted to use Claude for all purposes without restriction, but Anthropic refused to allow its use for large-scale surveillance or autonomous lethal weapons.

Subsequently, Anthropic released its top-tier models, Claude Mythos 5 and Fable 5, but the U.S. Department of Commerce imposed export controls banning foreign access to these models for national security reasons, only to lift the ban after 18 days. Anthropic is now working to improve its relationship with the U.S. administration by hiring a public sector executive with over 20 years of experience.

This content was produced with the assistance of AI translation services.