Bessent and Hassabis Both Call for
an AI Watchdog

In the third week of July 2026, two seemingly independent events converged: Treasury Secretary Scott Bessent helped draft a proposal for an independent AI agency reporting to the SEC, modeled on FINRA; and Google DeepMind CEO Demis Hassabis publicly called for a US-led international AI governance network targeting operational capability by the end of 2026. Two proposals from entirely different starting points — one financial stability, one AGI safety — but both arriving at the same conclusion: the current model of AI oversight is no longer sufficient.

This is not a coincidence.When a Treasury Secretary and the world's most visible AI lab CEO propose similar institutional frameworks in the same week, it signals that AI regulation is moving from "an academic discussion" to "a decision-maker action item."

Neither proposal is law — Bessent's plan is still under internal review, and Hassabis's call is not a policy document. But the direction is clear: AI needs independent, professional, enforceable public oversight — not voluntary industry commitments or fragmented self-regulation.

The Bessent Proposal: FINRA-Style AI Regulation

According to Bloomberg, Bessent's proposal calls for an independent AI agency modeled on FINRA (Financial Industry Regulatory Authority) — an industry-funded, SEC-overseen self-regulatory organization. The core function: pre-release safety screening of frontier AI models:

Why FINRA? Because it's a proven model. Formed in 2007 from the merger of NASD and NYSE regulatory functions, FINRA currently oversees over 3,400 broker-dealers and 624,000 registered representatives. For AI regulation — a domain requiring specialized expertise, rapid response, and industry funding — FINRA provides a directly referenceable institutional template.

Hassabis's Call: A US-Led AI Governance Network

The same week, DeepMind CEO Demis Hassabis issued a more urgent warning. His core argument: AGI is no longer a distant concept — it is "only a few years away." On this timeline, the world needs a US-led international AI governance network:

The differences matter too. Bessent's approach focuses on financial stability and national security — systemic risk from AI systems to financial markets. Hassabis's concern is AGI safety — longer-term, more fundamental existential risk. One pragmatic, one ambitious. Neither contradicts the other — together they span the spectrum of AI oversight.

China's AI Agent Recall Regulation: A Third Model

In the same week as Bessent and Hassabis made their calls, July 15, 2026, China's Implementation Opinions on AI Agents became legally enforceable — the world's first dedicated AI agent regulatory framework with recall authority. While the US is still discussing, China has already deployed enforceable regulation:

• Recall authority: Regulators can demand recall of non-compliant AI agents
• Three-tier decision authorization: High-risk agents require graded approval mechanisms
• Mandatory filing: High-risk sector agent deployments must be registered with regulators

Three models — Bessent's FINRA approach, Hassabis's international network, China's direct legislation — represent three paths for AI regulation. None is inherently superior, but they share one thing: they are all accelerating.

What This Means for Enterprises

Regardless of which regulatory path ultimately prevails, one trend is certain: AI regulation is moving from "soft guidance" to "hard enforcement." For organizations already deploying or planning to deploy AI systems, this means three things:

AI Governance Is No Longer a "Should We" Question

The third week of July 2026 may be remembered as a turning point for AI governance. Not because Bessent's proposal became law, or because Hassabis's call immediately produced an international agreement — but because the conversation itself has changed.

A year ago, the AI governance discussion was "enterprises should use AI responsibly." Now, the question has become "who should regulate AI," "what model should the regulator follow," and "can it be operational by the end of 2026?"

The shift in topic is itself the signal. When a Treasury Secretary and the world's most prominent AI scientist propose similar institutional frameworks simultaneously, AI governance is no longer something enterprises can "wait and see" about. It is becoming a fundamental constraint on how business operates.