The Age of Stupid—and Self-Serving—AI Regulation Is Here!
Anthropic and OpenAI may not like current federal controls on their products, but it will be consumers who end up getting screwed.
Back in 2018, Facebook founder Mark Zuckerberg minced no words as he was staring down congressional hearings over the regulation of social media and other parts of the online world. He didn't need the new AI-enhanced Meta Glasses he announced just yesterday to read the writing on the wall: A growing bipartisan consensus in Congress wanted Zuck—and the heads of Apple and Google and other tech giants—to do something about social media, which was starting to be blamed for everything bad in the world.
"I think the question is more what is the right regulation rather than 'yes or no should we be regulated?'" Zuckerberg told CNN. Previously a longtime critic of the need for government regulation, he started rethinking his position, and when he actually sat down before Congress in March 2018, he stated plainly, "We are ready to work with you to move beyond hearings and get started on real reform." To his credit, he didn't mince words that any and all regulations he was ready to help with would almost certainly benefit big players like Facebook: "When you add more rules that companies need to follow, that's something that a larger company like ours inherently just has the resources to go do, and that just might be harder for a smaller company getting started to be able to comply with."
Zuckerberg wasn't alone in his capitulation. By the end of 2018, Apple CEO Tim Cook was waving the white flag, too. "I am not a big fan of regulation," he told Axios. "But we have to admit when the free market is not working. And it hasn't worked here. I think it's inevitable that there will be some level of regulation….I think the Congress and the administration at some point will pass something."
It's hard to recapture exactly what wasn't "working" in the tech sector back then, but it was a mishmash of various topics and rumors-cum-conspiracies involving the mostly imagined throttling of some web traffic for nefarious reasons (thus necessitating the need for a brief and utterly useless experiment in Net Neutrality), unfair treatment of conservative commentators on Facebook and Twitter (does anyone else remember Diamond and Silk?), supposed subversion of American elections by Russians and North Koreans, and endless fights over Section 230 of the Communications Decency Act and arguing over whether it makes a distinction between platforms and publishers (it doesn't). In the end, a bunch of online platforms, many of them in the process of becoming woke (a term that originated in its current meaning in the 1930s but was only added to the Merriam-Webster Dictionary in 2017), banned hate-mongers like Alex Jones and Louis Farrakhan and Congress passed the Fight Online Sex Trafficking Act (FOSTA), which strangled aspects of online speech and commerce in the name of fighting sexual exploitation of minors. Federal regulators also went after companies like Google, the owner of YouTube, under an older, mostly dormant law called The Children's Online Privacy Protection Act (COPPA), generating what passed for big settlements related to purported failures to keep kids safe.
The specifics are less important than the very public turn among Big Tech firms toward what New York Mayor Zohran Mamdani would call the "warmth of collectivism," or government regulation.
I thought about that pro-regulation turn a lot while recently talking with Anthropic co-founder Jack Clark about the Trump administration's recent order banning the export of Claude Mythos and Fable 5, its most recent AI models, on national security grounds. The order, which may or may not be legal, follows the Pentagon labeling the company a "supply chain risk" after disputes about existing Defense Department contracts. (The designation means defense contractors can no longer use any of the company's products.)
Ironically, Anthropic has long called for state regulation of AI, and Clark's co-founder Dario Amodei published an essay just last week arguing in favor of "serious and binding regulation of AI." My full interview with Clark is here, but the 2-minute excerpt below explores why Clark doesn't think the AI industry is capable of self-regulation, even though his company voluntarily delayed releasing a product it deemed too powerful for general distribution.
Anthropic's pro-regulation sentiments echo an April proposal from OpenAI calling for governments to "implement common-sense AI regulation—not to entrench incumbents through regulatory capture but to protect children, mitigate national security risks, and encourage innovation."
Of course, of course! AI needs "common-sense" regulation, not the patently stupid sort that the Trump administration is currently pursuing against Anthropic, which reeks more of settling personal vendettas than any sort of systematic approach to the industry. And not the sort of ham-fisted interference envisioned by progressives like Sen. Bernie Sanders (I–Vt.), who has proposed a plan that would "give the public a 50% ownership stake in the largest AI companies in America."
But here is where it all gets confusing: On June 2, the White House issued an executive order about AI in the context of national security that was heavy on voluntary reviews of new products, a position at odds with its actions against Anthropic. And Vice President J.D. Vance recently said that, like Sanders, President Trump "is supportive of the United States owning these big AI companies." Democrats are falling over themselves to come up with ways to regulate AI, which in a recent national poll, "ranked less favorably than US Immigration and Customs Enforcement, President Donald Trump, former Vice President Kamala Harris, the Republican Party and The Late Show host Stephen Colbert."
You might also recall the claim by Netscape creator and venture capitalist Marc Andreessen that the reason he and other Silicon Valley bigwigs backed Donald Trump in 2024 was precisely because of Republicans' hands-off approach to AI—as opposed to Joe Biden's "seething contempt," in Andreessen's words, for the tech sector in general and AI in particular. (For what it's worth, Anthropic's Clark says the Biden administration wasn't particularly rough on the tech sector). As it stands, Meta, the parent company of Facebook, is "the only major U.S. developer of A.I. technology that has not reached an agreement to voluntarily share its models with the federal government for review."
When you've got virtually all of the big firms in an industry and leading politicians from all over the political spectrum calling for stricter oversight of a phenomenally unpopular technology (less popular than the GOP or Stephen Colbert!), it's a safe bet that whatever regulations are in place are only the start of what comes next. It's also a safe bet that those regulations, pushed by politicians who were born before America entered World War II (Sanders), who famously don't use email or even text (Trump), or who have deep ties to tech billionaires and trillionaires (Vance), will ultimately fix the market in favor of the firms currently doing well.
Remember what Zuckerberg said back in 2018 when looking down the barrel of regulation: "When you add more rules that companies need to follow, that's something that a larger company like ours inherently just has the resources to go do, and that just might be harder for a smaller company getting started to be able to comply with."