Washington's AI regulation fight splits between Zuckerberg's push for faster releases and Sanders' proposed ban on superintelligent systems, leaving policy in limbo.
Meta CEO Mark Zuckerberg told President Trump in a private August call that he opposes creating a national AI regulator. Senator Bernie Sanders, on the same Thursday, introduced legislation to ban advanced artificial intelligence outright. The two positions now define the extremes of a Washington fight that has no middle ground in sight.
The clash centers on how the US government should handle frontier AI models. Zuckerberg wants industry to move faster and police itself. Sanders wants the government to stop the most advanced systems from existing at all.
The regulator under discussion
Trump called Zuckerberg the week of August 17, according to Yahoo News. The president was weighing a proposal to create a watchdog modeled on the Financial Industry Regulatory Authority, or FINRA, which polices US brokerages and is funded by the firms it oversees.
The AI version would test frontier models for risk before release. Google DeepMind chief Demis Hassabis popularized the idea in July, framing it as a way to keep pace with rapid capability gains without stifling development.
Zuckerberg pushed back hard. He argued that a national regulator would slow American companies while foreign competitors race ahead. "Any policy that slows American model releases could add significant risk to American leadership while letting foreign models race ahead," he said in August. He had already staked out the position that superintelligence should reach everyone, not just a few well-funded labs.
The proposal is not dead. Officials are still weighing the FINRA-style body against a voluntary industry group. Trump adviser David Sacks favors the lighter option and has dismissed AI safety concerns as storytelling, which weakens the case for mandatory oversight.
The Sanders bill
Sanders and Representative Greg Casar announced the Ban Artificial Superintelligence Act on Thursday. The legislation would outlaw systems that match or exceed human cognitive performance.
That bar sits lower than the name suggests. Matching human performance on a given benchmark would trigger the ban, not some hypothetical far-future intelligence explosion. The bill would also freeze advanced AI work until a new federal regulator writes rules. Violators face up to 20 years in prison.
Sanders framed the stakes in stark terms. "The future of humanity cannot be left in the hands of a handful of Big Tech oligarchs," he said in a statement. He has pressed Congress on AI before without moving legislation, and the current bill faces long odds in a body that has struggled to pass even modest tech regulation.
What this means
The distance between these two positions reveals how fractured the AI policy debate has become. One side treats artificial intelligence as an economic race where slowing down equals losing. The other treats it as an existential threat where building faster equals losing. There is little appetite for the middle ground that Hassabis and others have tried to carve out.
The FINRA model has precedent in industries where self-regulation failed and government stepped in. But AI moves faster than brokerage markets, and a regulator funded by the labs it oversees would face the same capture concerns that have dogged FINRA itself. The voluntary approach Sacks prefers has no enforcement mechanism, which is precisely why Zuckerberg's camp likes it.
Meanwhile, the artificial intelligence index of deployed systems keeps climbing. Companies are shipping models with broader capabilities every quarter. Google, for instance, is rolling out a minimize button for its Gemini overlay so users can keep chatting while using other apps, a small but telling sign that AI is becoming ambient infrastructure rather than a discrete tool.
The policy gap matters because the next administration will inherit it. A voluntary framework can be dismantled with a memo. A FINRA-style body would require congressional authorization and years of staffing. Sanders' bill, even if it never passes, shifts the Overton window by making a total ban part of the conversation.
The real question is whether Washington can find a framework before the technology outruns the debate entirely. If Zuckerberg wins, the labs regulate themselves. If Sanders' framing gains traction, the political pressure shifts toward prohibition. Neither outcome addresses the harder problem: how to govern systems whose capabilities are still poorly understood by the people building them.
