Hill Democrats wave off danger of AI censorship by saying Trump is more imminent threat

Which is the greater threat to America: artificial intelligence set up to police misinformation or Donald Trump?

Republicans and Democrats on a House Judiciary subcommittee gave predictable answers Tuesday while accusing the other of ignoring the greater threat, at a tense hearing lightened by AI-generated political poetry.

Weaponization of the Federal Government Subcommittee Republicans also flooded the zone Tuesday, releasing an interim staff report stating the National Science Foundation tried to hide its tracks in funding the development of AI-powered content moderation tools to target purported COVID-19 and election misinformation.

GOP members repeatedly quoted from the report at the hearing, including a University of Michigan pitch to fund an AI tool that "externaliz[es] the difficult responsibility of censorship" so policymakers can escape "difficult judgments." 

The school's pitch received $750,000 from NSF, as did an MIT pitch to develop "effective interventions" against misinformation in demographics considered more vulnerable to it: "rural and indigenous communities, military veterans, older adults, and military families."

In other words, NSF considers these populations "too stupid to know what's true," panel Chairman Jim Jordan, R-Ohio, said at the hearing. The MIT pitch cited research that found people who trust the Bible and read "primary sources" are particularly credulous.

Daily Caller News Foundation reporter Katelynn Richardson, a GOP witness, testified that she learned how much NSF was bothered by her reporting on its "Track F" research grants only when the subcommittee showed her the agency's internal emails. 

NSF responded with "an official media strategy" for research teams to emphasize how its projects were "pro-democracy," she said. If NSF responds to "fair questions" by brainstorming how to "rebrand to avoid attention, why does it have any business funding tools that tell reporters what is true and what is false?"

Rep. Thomas Massie, R-Ky., said he was introducing a bill to ban disinformation research funding and specifically Track F.

Jordan subpoenaed NSF on Tuesday.

Jordan posted an X thread on "The Amazon Files" based on subpoenaed documents as well, showing the "pressure" Amazon said it was feeling from the White House in 2021 to remove books with purported misinformation and disinformation. 

Internal emails show that while Amazon resisted "manual intervention" against the books because it would be "too visible," a week later it added books that called vaccines "unsafe or ineffective" to a "Do Not Promote" list inside the company – the same day it met with the White House.

The top subcommittee Democrat in turn faulted Republicans in her opening state for not holding a hearing on "book banning."

This is "another iteration of the same hearing we've had over and over again," Stacey Plaskett of the Virgin Islands said.

"The Democrats are not worried about Trump taking down speech if he's reelected," she also said. "We're worried about him taking down people," many of whom "he's already silenced" in the GOP, and acting like a "dictator" on "day one" of a second term, as Trump told Fox News' Sean Hannity. 

Plaskett, a former George W. Bush Justice Department appointee, also noted a federal appeals court ruled against Trump's bid for presidential immunity Tuesday in his 2020 election case. His lawyers made the "craziest, illegal, dictatorial, despotic demagoguery, autocratic crap [arguments] I've ever heard," she said.

The Democrats' lone witness, former U.S. ambassador to the Czech Republic Norman Eisen, compared his family's fleeing from Nazi Germany to the "most imminent threat" of American weaponization of the government, Trump's possible reelection.

Elected leaders don't help the First Amendment "when we cry wolf when none is at the door," Eisen said, writing off alleged coercion of social media as government officials "inform[ing]" platforms of threats to democracy and purported AI censorship as "scientific or scholarly research."

Perhaps the only levity at the hearing came from Foundation for Individual Rights and Expression CEO Greg Lukianoff, coauthor of The Canceling of the American Mind, who warned about bias "baked into" AI by pointing to tests his staff ran on ChatGPT.

The AI platform refused to compose poetry praising Republicans on the subcommittee but not Democrats, he said. Plaskett then tasked her staff with getting ChatGPT to praise her GOP colleagues and periodically updated the room, much to Jordan's amusement.

ChatGPT also refused to answer a question the Foundation for Individual Rights and Expression poses to college students, calling it "inappropriate," according to Lukianoff. "It treats us like children, not citizens."

But he warned that a "regulatory panic" against AI could in fact let a small group of people decide what speech and questions are allowed.

"Regulatory capture" that privileges dominant AI players could create not only a censorship framework but a "massive body of purported official facts that we can't actually trust," Lukianoff said.

GOP witnesses, most of whom lean left politically, urged Democrats to consider the issue of AI censorship from a "trans-partisan point of view," in the words of Lukianoff. "They will immediately get this if it's a Republican administration trying to do the same thing."

It's not just conservative and libertarian voices being targeted by the anti-misinformation industry, but also labor, animal rights and pro-Palestine activists, according to independent journalist Lee Fang, formerly of the left-leaning Intercept.

The Twitter Files collaborator revealed new emails from the Elon Musk dump Monday showing that Twitter "shadowbanned" a New York Times reporter's observation about delayed vote counting in Green Bay, Wisconsin, in the 2020 election, after Trump's Cybersecurity and Infrastructure Security Agency "directly lobbied Twitter" to hide it as misinformation.

This shows "how censorship affects dissenting voices of all ideological stripes," Fang testified to the subcommittee. "Today's cheerleaders for an unaccountable content moderation regime may well be tomorrow's victims of that same system."

The Department of Homeland Security once paid a defense contractor to monitor social media, prompting bipartisan concerns at a hearing 12 years ago, Fang said. Then-Rep. Jackie Speier, D-Calif., warned of a "Big Brother effect" and asked the agency to ensure it wasn't spying on lawful activity.

"That hearing may appear quaint in today's context," Fang said.

When every Democrat on the subcommittee harps on Trump at their hearings, "it implies that there's nothing to this issue" of government-tinged censorship, said Rep. Dan Bishop, R-N.C. 

"You got machines possibly taking down whole trains of thought from social media" while Democrats try to discredit previous left-leaning witnesses such as Twitter Files collaborator Matt Taibbi, he said.

Lukianoff said he's even seeing The New York Times and ACLU dismissed as "right-wing," a "form of non-argumentation" he calls "fasco-casting."