The current landscape of artificial intelligence policy in the United States presents a paradox, according to Microsoft President Brad Smith, who recently characterized the situation as “regulation without transparent or complete rules.” This assessment, delivered during the AI for Good Global Summit, highlights a growing apprehension within the tech industry regarding the federal government’s ad hoc interventions in the development and deployment of advanced AI models. Companies, Smith suggests, are struggling to navigate an environment where clear guidelines remain elusive, hindering strategic planning and innovation.
Recent actions by the Trump administration have underscored this ambiguity. The Commerce Department, for instance, utilized export-control laws to compel Anthropic to withdraw its Fable 5 and Mythos 5 models from global markets, citing cybersecurity risks. Weeks later, OpenAI reportedly faced pressure to delay the public release of its GPT-5.6 model family, with initial access limited to government-approved partners. While these restrictions have since eased — Fable 5 is back online, and GPT-5.6 is slated for public launch — the incidents have left AI developers uncertain about the government’s overarching policy framework. Smith acknowledged the government’s right to act on genuine security concerns, particularly in the case of Anthropic’s Fable, noting that frontier models do require pre-release assessment. However, he pointed out that the government resorted to an export control tool, which legal experts contend was not originally designed for widely accessible AI models delivered via API.
This reliance on an ill-fitting regulatory instrument suggests a fundamental deficit in the government’s toolkit for managing emerging AI technologies. Smith observed that the U.S. government, despite having information about an urgent cybersecurity risk, found itself with only one available regulatory lever. This approach, critics argue, resembles an informal licensing regime operating without formal legislation or clearly defined criteria. A June executive order had established a voluntary pre-release review process, deliberately avoiding a mandatory licensing system. Yet, the subsequent actions against Anthropic and OpenAI indicate a willingness to employ mandatory measures when voluntary cooperation is not secured or when perceived risks are high, all without a published standard for intervention or criteria for identifying “trusted partners.”
The government’s use of export controls, specifically in blocking foreign access to Anthropic’s models, has also instigated a scramble for “sovereign AI” globally. This concept emphasizes national control over AI models and their supporting infrastructure. European politicians, in particular, voiced strong reactions, interpreting the U.S. move as a stark illustration of the dangers inherent in over-reliance on American technology. One French official compared the shutdown to a blockade, while a British lawmaker highlighted the immediate impact on hospitals and researchers who lost access to critical technology. Canadian Prime Minister Mark Carney echoed similar sentiments, emphasizing the perils of dependence on a limited number of providers.
However, Smith believes the export control action against Anthropic has been largely misinterpreted as an attempt to specifically cut off foreign users. He clarified that the government’s intention was to compel Anthropic to remove Fable from the market entirely, both domestically and internationally, after the company initially declined to do so voluntarily. Regardless of the intent, foreign governments now view the Anthropic episode as evidence of their vulnerability to U.S. infrastructure. Smith contends that the onus is now on Washington and American tech firms to reassure the global community about the reliability and continuity of access to their systems. He stressed that international customers will only invest in U.S. services if they are confident in the certainty of supply, emphasizing that “there needs to be a level of technology assurance, both access to markets and access to supply. The two go together.” This sentiment resonates with broader industry concerns about maintaining global trust and market access in a rapidly evolving technological landscape.
