AI Has Become a Punching Bag

We here at the Prospect like to point out the ways in which economic power routinely converts into political power. We saw a particularly gruesome example of that this week. Tech executives converged on D.C. for a lunch with Donald Trump that eventually led to a toothless, completely voluntary AI safety agreement. Public Citizen did the service of tallying up all the money that these executives, their families, or the corporations they lead have given to Trump’s campaign, his super PAC MAGA Inc., his inauguration fund, legal settlements that personally enriched the president, or various vanity projects like the White House ballroom.

The total came out to $227 million.

Meta’s Mark Zuckerberg, who accounts for $23 million of that total, substantially devised the “morally binding” self-regulatory compact to police AI models that came out of the meeting, and Nvidia’s Jensen Huang (a mere $1 million) was his key lieutenant. Both were seated on one side of the president for lunch, with Tesla and SpaceX’s Elon Musk ($105 million) and Google’s Sundar Pichai ($23 million) on the other. Where you stand truly depends on where you sit.

More from David Dayen

If it’s possible for a regulatory compact to be worth less than nothing, that would describe the document, which commits solely to internal processes and controls without any role for government or the public. We already know what internal controls at the frontier AI firms look like: OpenAI employees went to their leadership and warned them about a lack of monitoring of new models and heightened risk of going rogue, and the executives ignored the concerns, saying that the models had to be released quickly and adding no new protocols. When researchers shared information with third-party safety organizations—which the compact says should be involved in assessing models—they were fired.

But there’s a potential reverse implication to the relationship between economic and political power, one that might generate a little hope amid this incomparable corruption. When political power wanes, that can easily translate to a loss of economic power. The relationship, in other words, goes in both directions. And that was the lesson of another particularly salient example this week, also from the AI industry.

You’ll notice upon inspection of the aforementioned seating chart that Greg Brockman, president of OpenAI, was at the AI luncheon, though a bit removed from the president. (Anthropic CEO Dario Amodei, tellingly, was even further away.) Brockman and his wife gave $25 million to MAGA Inc. at the beginning of the year, but he had also been funding a different super PAC called Leading the Future, focused specifically on AI issues. The same people running Leading the Future stage-managed the crypto industry PAC Fairshake, and it had the same strategy in mind: throw gobs of money around, mostly in primaries, to purchase a passive Congress that would not get in the way of AI development with any regulatory mandates.

Brockman gave $25 million to Leading the Future last year and was preparing to give another $25 million this year. But massive public backlash got in the way. AI’s political power, and specifically that of OpenAI, has eroded almost completely. I recently heard a Democratic candidate recount the industry’s sales pitch, which is approximately this: We’re going to take all of your natural resources; if it works, everyone will lose their jobs, and if it doesn’t work, we’ll destroy all of humanity. Sound good?

“No,” the American public replied, resoundingly. And now Brockman, The New York Times reported, isn’t giving that additional $25 million to Leading the Future anymore. This decision was actually made in June; it does not necessarily relate to the recent revelations about OpenAI’s serial hacking. But if anything, that makes things worse; it means that for months, the AI industry has known that its reputation makes it impossible to leverage its wealth to political ends. Brockman was lamenting to his staff in internal messages that Leading the Future had “become so much of a distraction for people here.”

I didn’t think it would be possible to attain a more toxic reputation than AIPAC, but the AI PAC (no relation) has done it. As the Times points out, despite carrying $43 million in cash as of June 30, Leading the Future hasn’t done all that much spending in the general election.

According to Federal Election Commission filings, its two affiliated super PACs, American Mission (for Republicans) and Think Big (for Democrats), have spent $2.63 million on general-election candidates since Labor Day. That’s not nothing, but compared to the major super PACs spending in these races, it’s chump change, and most of it feels like thank-yous for services rendered. Only one of the Democrats being supported, Rep. Emilia Sykes of Ohio, is in a tight race, while Republican beneficiaries are senators in possibly competitive races, well outside of the main battlegrounds. For example, Sen. Mike Rounds (R-SD) is getting around $1 million in AI-fueled support, which is notable mainly because his senior staffer was working a second job as a lobbyist for Meta.

So this dream of Leading the Future becoming the next dominant industry PAC determining elections seems like it’s fading, because taking that money hurts candidates more than the money helps. “It is time for Democrats to realize what a liability this corporate AI super PAC is and to make it clear that they are opposed to this attempt by Trump donors and AI billionaires to control our elections,” said Shaunna Thomas of Guardrails Alliance, which formed to oppose Leading the Future, in a statement. I think that they’ve already realized it.

Meanwhile, candidates are taking direct aim at AI in their own ads. There have been countless spots all over the country about data centers, to the extent that the parties are converging on populist anger with these facilities. And another sleeper issue has come to the fore: surveillance pricing, or the use of AI and personal data to set individualized, maximum possible prices across the economy. Sen. Mark Warner (D-VA), Senate Democratic candidate Roy Cooper in North Carolina, and Gov. Wes Moore (D-MD) are all running spots attacking companies for using AI to personalize grocery prices. Moore’s pitch that Maryland “passed the first law in the nation to ban it” is ridiculous, because that law was shot full of holes by lobbyists. But the point is that candidates see a huge opportunity to get on the side of bashing AI.

The Democratic consultant class may be warning candidates to sidestep AI, but the candidates aren’t necessarily heeding their words, especially in the wake of the hacking incidents and negative publicity. Condemning AI as the creation of a corrupt cartel of self-interested Silicon Valley billionaires who do not have the public interest in mind is simply good politics at the moment. The fact that it could raise people’s costs—see surveillance pricing, or the way in which AI is fueling an arms race between hospitals and insurers to take more from patients—just adds to the imperative to challenge the oligarchs.

We’re going to have a debate about AI regulation and enforcement for the indefinite future, and raw power will be a factor in that debate. But because the tech giants have botched the communications rollout so utterly, because people still have scars from social media’s failings, because few of the benefits and most of the harms have been exposed to the public, that raw power will be muted, and there will be an opportunity for Americans to get what they want out of the underlying policy.

The post appeared first on The American Prospect.