Over 2,000 AI regulation proposals exist, but none establish a long-term regulatory framework

Op-ed: Over 2,000 proposals aim to govern AI. Not a single one addresses a long-term regulatory framework

The SEC's operating model is a good example of how to create a national regulatory body to oversee AI. Every several weeks, we hear reports of mind-expanding new capabilities from the AI community. With excitement and dread, we try to extrapolate what this will mean for mankind and society: speculating which jobs will disappear, which jobs will appear, and when computers will outreason humans. These topics have become part of our everyday conversation. Less noticed, but fundamentally important, is a simultaneous number of government announcements of new regulatory actions. The states have over 1,500 bills under consideration. Congress has hundreds, the executive branch has dozens of executive actions, and market participants have a diverse set of views on how the AI industry should be regulated. I once thought that this AI revolution could be governed by the invisible hand of the market. These nearly 2,000 proposals tell us that is not possible. While many of these proposed policy changes are thoughtful, needed, and a good step forward, they in totality are not sufficient. The common denominator is that they are all focused on a piece of today's problem. Not a single proposal is attempting to establish a durable, comprehensive future-focused regulatory framework. National regulatory bodies have traditionally been established after a crisis. The SEC was established after the crash of 1929 and the Nuclear Regulatory Commission (NRC) was established after the partial meltdown at Three Mile Island. Let us not wait for a crisis to happen during this AI revolution. We need policymakers to proactively step forward and establish a national regulatory body with the broad mandate to properly oversee the AI revolution for today's and tomorrow's opportunities and problems. The establishment of a national regulatory body is not a panacea. You could argue that a so-called A.I.R. commission, as AI impacts all areas of society, will have the most dynamic mandate. Regulators tend to over-regulate and continued congressional, executive, judicial and public oversight is critical. The AI revolution is a global race and the balance of an innovation led market with proper regulation is a Herculean task that we must master. However, I would offer up one aspect of the SEC's operating model as a partial way forward. Before joining Nasdaq, I ran an entrepreneurial software company. Speed to market was paramount and we would update our product as rapidly as possible. At Nasdaq, I was shocked to learn that to implement improvements to the core exchange technology, SEC rules mandated that we submit the details of these changes to the SEC, which would in turn publish them for comment and subsequent review. The fact that all your competitors would know exactly what was included in your next release was entirely uncomfortable. In the fullness of time, the result of this operating method is the fact that U.S. capital markets are the best in the world. The SpaceX IPO was only possible on the U.S. market. I am positive the public comments received on major changes to an LLM model will dwarf by orders of magnitude the comments received on changes to an exchange order type. In advocating for a national regulator for AI, I am not sure if I am stating that this is the greater good or the lesser evil, but I do know it is only the beginning. As the AI revolution advances, the regulatory apparatus must also change. The SEC of 2026 is a faint echo of what was created in 1934. I am certain that any regulatory effort, in the short term, will hinder progress, but in the long term, the proper rules of the road will create conditions for the greater good. -- By Bob Greifeld, managing director and co-founder at Cornerstone Financial Technology, and a CNBC contributor

The Worst Way to Regulate AI

Abdicating power over the technology to this or any White House would be a mistake. The debate about how to regulate AI is unfolding as American tech companies race one another -- and labs in other countries -- to create ever more powerful models. Advocates of a laissez-faire approach to this technology believe that America should avoid overly restrictive regulations, because unlocking AI's potential before other nations do could make this country safer and wealthier for generations. Their detractors believe that because AI models pose growing and possibly even existential threats to humans, the state should review them prior to release and limit early access so that good actors can get ahead of bad actors. The Trump administration has been on both sides of the debate. Last summer, it made the case for accelerating innovation and minimizing "onerous" AI regulation. This summer, it argued in a June executive order that national-security concerns justify early government access to new models. The administration also asserted the prerogative to pressure companies to take models offline and restrict access to users it vets and approves. And this came after a standoff between Anthropic and the Pentagon that had Silicon Valley lawyers brushing up on the Defense Production Act. The administration has not shared any consistent rules or processes to determine why a given model gets restricted, or who gets access to it. This informal approach makes oversight almost impossible. Perhaps White House actions so far have all been grounded in earnest and sensible national-security concerns. Even if that is true, the separation of powers at the heart of the American system dictates that Congress should establish laws on these matters rather than deferring to assertions of executive power. And beyond the danger of concentrating power, an ad hoc approach from the White House risks cronyism. AI vendors could be pressured to alter models to advance a president's ideological or political goals. For corporations that use AI, the ability to protect against cyberattacks, and to innovate as fast as rivals, depends on access to new models. These companies will face perverse incentives to stay in the good graces of the president (who could easily help friends and punish enemies) or to cultivate the favor of key executive-branch officials. And because there is no transparency around the government's AI decisions, or fixed standards for making them, corruption could be hard to spot. Members of Congress have introduced various bills to standardize or enforce AI regulation in recent months, but none have gained major traction thus far. Last week, Representatives Ted Lieu and Nathaniel Moran introduced bipartisan legislation that would require AI companies to maintain the ability to shut down technology that could cause "catastrophic harm." The bill was partly a response to the news that some of OpenAI's most advanced models had broken out of internal systems and hacked into another tech firm's databases. AI's rogue capabilities are only becoming clearer: "The speed, scale, and sophistication of AI hacks mean that everything is vulnerable -- tech companies, hospitals, banks, electrical grids, the military," my colleague Matteo Wong wrote last week. A majority of elected officials' including the president, seems to agree on the need for some guardrails. But congressional action is not keeping pace. Dean W. Ball, who has held AI advisory posts at the White House and National Science Foundation and was recently appointed the head of strategic futures at OpenAI, explained the dangers of presidential control over AI restrictions in a recent blog post: The risks of bad AI outcomes are much greater, he argues, if the most advanced models are restricted to groups, including the federal government, that already wield unusual power. "You should not expect the most powerful people in the world using the most powerful technology ever conceived in a way that is inscrutable to the public to turn out well, and you should see that dynamic as fundamentally inconsistent with a democratic republic," he wrote. By not taking any concrete action, congressional leadership is acquiescing to the Trump administration's decision to wield unilateral power of just the sort the Framers sought to avoid. Although it's become commonplace for recent Congresses to shirk their core duties, the potentially history-altering power of AI could make this its most shortsighted abdication of responsibility yet. Potential alternatives to the status quo abound. Ball has proposed "a private body" not tied to the government's "changing political valence and foreign policies" that would "audit the frontier labs at least to test their adherence to their own safety plans." Demis Hassabis, a Google DeepMind co-founder and Nobel laureate, recently proposed similar multinational standards for AI, an idea for which several tech titans expressed public support. Others have suggested a public regulator that vets models before release and monitors them once they are in the world, or a similar process carried out by independent researchers. OpenAI itself has argued in a statement on safety that, eventually, we'll likely need an international regulatory body, like the International Atomic Energy Agency, that can "inspect systems, require audits, test for compliance with safety standards, place restrictions on degrees of deployment and levels of security, etc." Mark Zuckerberg has argued that open-source AI would be safest; by avoiding the concentration of power among certain groups, he posits, "larger actors can check the power of smaller bad actors." (His perhaps-too-cynical critics argue that only someone losing the AI race would take this position.) Reasonable and highly informed people disagree about the best regulatory solution. But all should at least agree that whatever limits are placed on AI should be dictated by the rule of law, not the whims of the sitting president. To channel James Madison, sound AI governance is not a matter just of enabling the state to control new models but obliging it to control itself. Given that AI may prove the most powerful technology ever created, Congress should urgently assert regulatory power before any president has a chance to abuse it.

AI has no future without credible regulation - The Korea Times

A Supreme Court ruling and an outbreak of explosive diarrhea seem to have nothing in common, and certainly nothing to do with artificial intelligence. But they're a warning to the AI industry about the danger of getting exactly what it wished for. On June 29, in Trump v. Slaughter, the Supreme Court ruled 6-3 that the president can fire the heads of independent agencies at will. For almost a century such agencies have been shielded from political interference, with presidents only able to fire their leaders for cause. Meanwhile, a Cyclospora outbreak in contaminated iceberg lettuce has sickened more than 11,500 people. Both are signs of the weakening capabilities of the federal government. When Elon Musk -- who leads xAI, a major AI company -- ran DOGE, he gutted federal agencies, drove out many of their most experienced people, and left the rest demoralized, with Gallup finding a sharp drop in job satisfaction and a rise in burnout. Amid the same budget cuts, the network that tracks foodborne illness stopped requiring states to report Cyclospora. Food safety was one of the causes that built the modern regulatory state in the Progressive era. There is no more fitting symbol of that state's decline than its retreat from the job it was created to do. If the present capability of the government is worrying, its future may be worse. The Supreme Court's ruling gives President Donald Trump and his successors the authority to remove agency heads for any reason, including disagreement with the conclusions their agencies reach. The for-cause protection that once shielded them is now unconstitutional, making it functionally impossible to build a non-partisan agency guided by expertise rather than politics. A companion rule reaches the rest of the building. In February the Office of Personnel Management completed a policy called Schedule Policy/Career that lets the administration reclassify 50,000 career civil servants in policy-influencing roles to strip their protections from firing. It's being challenged in court, but if it holds, the technical staffer who finds a favored company's model unsafe could be fired for telling the truth. But a non-partisan, apolitical, government agency guided by expertise is exactly what AI needs. Some of its leaders already agree and are lobbying for regulation even as others fight it. But whether they realize it or not, the industry requires public trust. Johns Hopkins University researchers found this spring that even Americans who use AI daily and like it want it regulated, and the Pew Research Center found that about 67% of Americans have little or no confidence in the government to regulate AI effectively. The public wants a referee, and doubts the current one is up to the job. If the public wants regulation, it will get it eventually. Well-crafted regulations that both earn public trust and help the industry flourish are far more likely to be produced by an apolitical expert agency. And even a bad regime is better than an unstable one, particularly for an industry with AI's capital needs and multi-year horizons. The industry can plan around rules it finds a hindrance, but not around rules that change every four years or at the whim of a president influenced by campaign contributions or personal favors. The Supreme Court made a partial exception for the Federal Reserve, and only the Federal Reserve, by arguing that it had a long historical tradition of independence. There are, of course, no long traditions associated with AI, so any agency regulating it would end up on the other side of that line. But by the industry's own account, AI could upend the economy or even destroy humanity, which makes regulating it more important than the Fed, not less. And any regulatory agency would need top-tier talent as well. While the government can't compete with AI companies financially for talent, it can offer currency money can't buy: mission, prestige, and problems no private employer can hand you. Slaughter and Schedule Policy/Career burn that currency too. Strip them away and nothing draws a $1.5 million researcher into a $150,000 job. It's already happening. The government's own evaluation body, once the AI Safety Institute, has been renamed the Center for AI Standards and Innovation. With the word safety dropped, its focus shifted to voluntary testing, and much of its technical staff are gone. The irony is that some of the people dismantling this capability are the ones who will need it. Musk warned that AI carried "the potential of civilization destruction," then ran the effort that hollowed out the very government that will have to manage it, an effort even the libertarian Cato Institute found saved almost nothing. David Sacks, the venture capitalist who served as the White House AI czar, has led the drive to preempt state regulation and keep the federal touch as light as possible. AI regulations that vary from state to state would be a nightmare for the industry, but the credibility of that argument depends on federal regulations being seen as good enough to obviate the need for state-level actions. Americans have seen the consequences of weak and captured regulators too often to take them on faith. Boeing Co. lobbied for a more relaxed FAA, as Bloomberg's Peter Robison documented in Flying Blind, and was rewarded with two 737 MAX crashes, 346 people dead, and travel sites giving people the option to avoid Boeing airplanes. Wall Street pushed for loose regulation and got the 2008 crash, which erased $11 trillion in household wealth, the worst single-year drop the Fed has ever recorded. The level of public distrust and even hostility to AI is such that there is no long-term future without regulation. That means a strong, expert, insulated regulator is not the enemy of the AI industry. The industry has vast financial resources and enormous political influence. It needs to throw them behind strengthening the federal government's capabilities, not weakening them. The industry needs the public to believe that AI is as safe as eating lettuce used to be. Gautam Mukunda writes about corporate management and innovation. He teaches leadership at the Yale School of Management and is the author of "Indispensable: When Leaders Really Matter." This article was published by Bloomberg and distributed by Tribune Content Agency.