DeepMind CEO Proposes FINRA-Style AI Regulator as AGI "5 Years Away" Warning Intensifies

Google DeepMind CEO Demis Hassabis has proposed a FINRA-model regulatory framework for AI, featuring industry-funded oversight with government supervision and technical experts to conduct pre-deployment safety testing on frontier models by year-end.

Speaking at the 2026 Silicon Valley Roundup, Hassabis warned that artificial general intelligence could arrive within the next five years, significantly accelerating the timeline for needed safeguards.

The DeepMind chief also cautioned that "dangerous capabilities" could appear in open-source models within 18 months, raising concerns about proliferation beyond the control of major AI labs.

"AGI could be here in 5 years. We need a regulatory framework that can move as fast as the technology itself." – Demis Hassabis, CEO of Google DeepMind

The proposal draws inspiration from FINRA, the Financial Industry Regulatory Authority, which operates as a self-regulatory organization with government oversight, funded by the industry it regulates.

Under Hassabis's vision, an AI regulatory body would be financed by the major frontier model developers but operate independently with technical expertise and government supervision.

The goal is to create a framework that can conduct pre-deployment safety testing, enforce transparency standards, and respond quickly to emerging threats—all without requiring new legislation that could take years to pass.

Hassabis's proposal comes amid growing urgency across multiple fronts. More than 200 economists and researchers, including 15 Nobel laureates and scientists from OpenAI, Anthropic, and Google, have issued a joint warning that AI's economic transformation could happen in years rather than decades, calling for urgent policy action.

Meanwhile, China has already released its AI Safety Governance Framework 2.0, setting standards for risk classification, transparency obligations, and safety testing requirements.

The framework includes mandatory safety assessment and registration for models above a certain capability threshold, with violators facing fines of up to 7% of global annual revenue.

As the timeline for AGI continues to narrow, the gap between technical capability and governance structures becomes increasingly concerning.

The next 12-18 months may determine whether the industry can establish effective self-regulation—or whether governments will step in with more prescriptive measures.