Analysis
Anthropic CEO Dario Amodei pushed back over the weekend against criticism that his own public messaging has fueled the growing backlash against AI, while simultaneously conceding the industry has fallen short of its own promises, according to Fortune. Writing that AI companies 'haven't yet delivered on our big promises to benefit the world,' Amodei framed public skepticism as a rational response to a long history of institutions and companies using new technology against ordinary people's interests -- and endorsed the creation of a FINRA-style independent regulator for frontier AI companies, an idea Google DeepMind CEO Demis Hassabis proposed earlier this year.
The alignment between Amodei and Hassabis -- executives who run directly competing labs, Anthropic's Claude against DeepMind's Gemini -- on a specific regulatory structure is itself notable. Amodei explicitly called the moment 'a meaningful change compared to six months ago, when most of the industry was only calling for the repeal of state regulations with no clear federal approach,' a description that fits Anthropic's own public position for much of the past year, when it lobbied against a broad federal preemption of state AI laws while OpenAI and others pushed harder for deregulation.
“A FINRA-style body -- modeled on the Financial Industry Regulatory Authority, the self-regulatory organization that oversees U.S.”
A FINRA-style body -- modeled on the Financial Industry Regulatory Authority, the self-regulatory organization that oversees U.S. broker-dealers -- would function as industry-funded but independently governed oversight, sitting between pure self-regulation and a traditional federal agency like the SEC. Applied to AI, the model would likely require frontier labs to submit models for testing against safety benchmarks before or shortly after deployment, with enforcement authority the current patchwork of state laws and voluntary lab commitments doesn't provide -- a meaningfully different proposition than the model-card disclosures and red-teaming pledges labs have relied on to date.
The timing matters given where Anthropic sits commercially. The company's annualized revenue run rate reportedly hit $65 billion in July, and investors are modeling a prospective IPO valuation between $2 trillion and $3 trillion for as soon as October -- a scale that puts Anthropic's own regulatory posture under public-company-level scrutiny well before it actually lists. A CEO who has spent the past two years warning publicly about AI risk while running one of the two or three companies best positioned to profit from AI's growth has an unusually direct stake in what any resulting regulatory framework actually requires.
The bear case on Amodei's framing: a FINRA-style body, funded and staffed substantially by the industry it oversees, has historically been criticized in financial services for regulatory capture -- FINRA itself has faced years of criticism that broker-dealers wield outsized influence over its rulemaking. An AI equivalent could end up formalizing the same self-interested standard-setting critics already accuse frontier labs of doing informally through voluntary safety commitments, just with a more official-sounding name attached. OpenAI, xAI and Meta have not publicly endorsed the same proposal, and without buy-in from every major lab, a FINRA-style body covering only Anthropic and DeepMind's voluntary participants would carry far less weight than the two CEOs' public alignment suggests.