Martin Chavez, a member of Alphabet’s board and vice chairman of investment firm Sixth Street, sharply criticized the current U.S. approach to artificial intelligence regulation as inconsistent and unclear for developers and companies. Speaking at a Reuters event in London, Chavez highlighted the fragmented system where each AI model undergoes individual assessments with little transparency about decision-making processes.
Chavez argued that this reactive, case-by-case scrutiny creates an unpredictable environment that hinders innovation. He suggested regulators adopt a framework similar to the stress tests imposed on banks following the financial crisis—periodic, standardized evaluations designed to identify systemic risks after real-world failures surface. This approach, he said, would better align oversight with the fast-evolving nature of AI technology.
The concern over AI’s rapid growth and its economic impact is growing among central bankers as well. The Bank for International Settlements cautioned in its 2026 Annual Economic Report that the AI investment boom faces pressure and risks including workforce disruption and supply chain bottlenecks. Policymakers are also increasingly factoring national security into AI governance.
In recent months, export control measures have intensified. Anthropic, a major AI developer, announced it would disable some of its most advanced models following a U.S. government directive restricting access by foreign nationals. The company opposed withdrawing widely deployed commercial products but faced mounting pressure amid vague national security justifications. Similarly, OpenAI delayed the full public launch of GPT-5 under governmental request, initially restricting use to preapproved partners whose information was shared with authorities. These steps demonstrate how regulatory bodies are shaping AI deployment even before comprehensive legislation is established.
Chavez’s critique also recalled prior tensions between Anthropic and U.S. military agencies over the use of AI models, which culminated in the company’s placement on a supply-chain blacklist effective later in 2026. This reflects an AI oversight landscape emerging in disjointed fragments—combining individual model reviews, export restrictions, and security-driven delays.
Companies developing AI now face navigating a complex maze of regulatory demands, national security concerns, and investor expectations without a unified framework. Chavez’s call for stress-test style oversight aims to address these challenges by fostering clearer, outcome-based regulation that adapts as AI technology and its risks mature.