Martin Chavez, vice chairman of investment firm Sixth Street and a member of the board of Alphabet Inc., said the United States' current approach to regulating artificial intelligence is flawed and lacks consistency. Speaking at the Momentum AI London event, Chavez criticised the practice of assessing each new AI model's release individually, saying it produces insufficient transparency about who is making regulatory decisions and how those determinations are reached.
Chavez framed his concerns with a reference to past regulatory cycles in finance, saying that - "The right regulation always happens after the fact." He linked that pattern to a reluctance to act before harm materialises, arguing that authorities frequently wait for evident negative outcomes before imposing rules.
As he addressed the conference, Chavez urged consideration of a more systematic safety framework for AI. He suggested that oversight could mirror the post-financial crisis regime that introduced annual stress testing for banks, with the aim of creating a baseline of safety across the sector rather than ad hoc, model-level interventions.
His comments came as other organisations and regulators have expressed concerns about the scale and speed of AI deployment. The Bank for International Settlements warned recently about potential consequences of the AI boom, including job losses, supply bottlenecks and a pattern of excessive investment that in previous boom-bust cycles produced substantial losses.
Chavez's remarks also come against a backdrop of concrete government requests to AI companies. On June 12, Anthropic said it would disable its most advanced AI models for all users after the U.S. government ordered it to suspend access for foreign nationals on national security grounds. On June 26, OpenAI announced it was postponing a full public launch of GPT-5.6 at the U.S. government's request, providing initial access only to a small, vetted group of partners whose identities were shared with authorities.
Those developments illustrate the immediacy of regulatory pressure and the uneven ways governments are interacting with AI firms. Chavez said the industry would benefit from overarching safety requirements that apply broadly, rather than a sequence of bespoke interventions tied to individual model releases.
The substance of Chavez's critique focuses on transparency and the need for durable safeguards. He warned that the present pattern - individualised actions and limited disclosure about decision-making - could leave gaps in oversight and uneven protections across different models and firms.
Summary: At Momentum AI London, Martin Chavez criticised the U.S. practice of evaluating AI models on a case-by-case basis, called for more transparent decision-making, and recommended systemic safety measures similar to post-crisis financial stress testing. Recent government actions affecting Anthropic and OpenAI underscore mounting regulatory scrutiny.
- Key points:
- The U.S. approach of regulating individual AI model releases is inconsistent and opaque, according to Chavez.
- Chavez advocates for sector-wide safety requirements comparable to financial stress tests to raise baseline protections.
- Recent government interventions - Anthropic disabling advanced models on June 12 and OpenAI delaying GPT-5.6's full public launch on June 26 - demonstrate active regulatory engagement.
- Risks and uncertainties:
- Regulatory inconsistency and a lack of transparency could increase systemic risk in the AI sector, affecting technology companies and investors.
- The AI boom raises concerns about job displacement and supply bottlenecks, posing risks to labour markets and supply chains, as noted by the Bank for International Settlements.
- Rapid, concentrated investment in AI could recreate boom-bust dynamics that produce heavy losses for investors and related financial markets.