A newsletter briefing on the intersection of technology and politics.

At a hearing whose star witness was Dorota Mani, the mother of a New Jersey teen allegedly victimized by schoolmates’ circulation of faked nude images, there was no debate that deepfake porn and nudes are a problem — one that disproportionately affects women and children and can ruin their lives. In his opening remarks, Rep. Gerry Connolly (D-Va.) cited a recent report that claimed 98 percent of all online deepfake videos were pornographic and that women or girls were the subject in 99 percent of them. 

The debate now is what Congress should do about it — and what it can do without running afoul of the First Amendment.

In one corner of the debate was Carl Szabo, vice president and general counsel of NetChoice, a tech industry group funded by Google and Meta, among others. Testifying at the hearing, Szabo argued that most of the harms posed by AI deepfakes can and should be addressed through existing laws rather than new ones.

“Every law that applies offline applies online,” he said. “So when it comes to harassment, we need to enforce harassment law. When it comes to fraud, we need to enforce fraud law.”

Szabo said there are some gaps in existing laws when it comes to deepfakes, however, that could be addressed with narrowly tailored legislation. For instance, he said NetChoice supports an amendment to clarify that artificially generated child pornography violates the same laws that apply to other forms of child sex abuse material. But he said he's wary of government “overreach,” adding: “The last thing we want is to make a law that doesn't hold up in court.”

His testimony before the House Oversight cybersecurity subcommittee signaled that the tech industry will fight legislation that seeks to hold tech firms responsible for AI-related harms. It will push instead to put the onus on individual “bad actors” who abuse the technology.

The problem is that “current laws aren’t working,” countered Ari Ezra Waldman, a law professor at the University of California at Irvine. While it has always been possible to manipulate images, he said AI presents a “proliferation problem,” because it has dramatically lowered the barriers to both creating and sharing those images. 

He argued that both civil and criminal penalties are needed to deter deepfake porn, and that platforms should bear some of the responsibility. Suing the individuals who create and publish the material isn’t enough to make victims whole, he said, because “so many of the perpetrators of this are the dude in the basement who's probably judgment-proof.”

Relying on AI companies to police abuse of their own software is another dead end, argued John Shehan, senior vice president of the National Center for Missing and Exploited Children. 

As reports of AI-generated child exploitation material roll in to NCMEC's CyberTipline, the majority are not coming from the AI companies themselves, he said. In fact, many of the companies behind prominent AI image tools haven't even registered to make such reports. 

Then there is the problem of distribution. Connolly pointed to the Taylor Swift deepfake nudes that circulated widely on X in January, saying the fact that it happened to a star with as much power as Swift “emphasizes that no one is safe.” And my colleague Drew Harwell reported last month that deepfakes of online influencer Bobbi Althoff, which first surfaced on porn sites, quickly went viral when they were posted to X.

Social media companies such as X have been able to lean on Section 230, an immunity shield for websites that host user-generated content, for protection against lawsuits stemming from harmful material their users post. There's an emerging consensus that Section 230 may not protect tech firms from lawsuits arising from material generated by their own AI tools, but plaintiffs still need a cause of action to win such suits.

There are House bills in the works to address the issue, though they face a long road to passage.

One that has Mani’s backing, and some bipartisan support, comes from Rep. Joseph Morelle (D-N.Y.). His bill would criminalize the sharing of nonconsensual, digitally altered images and create a private right of action for victims of it. And the subcommittee’s chair, Rep. Nancy Mace (R-S.C.), introduced a deepfake pornography bill of her own earlier this month, which so far has three Republican co-sponsors. 

Waldman told the subcommittee Tuesday that the First Amendment shouldn't stand in the way of deepfake laws, which he compared to existing laws prohibiting counterfeiting, impersonation and forgery, as long as they carve out exceptions for satire and other protected speech.

Ari Cohn, free speech counsel at the digital rights group TechFreedom, told The Tech 202 he isn't so sure. While Congress may be able to show a compelling interest in regulating certain types of deepfakes, he said, “not every deepfake is inherently defamatory or fraudulent,” and “any law treating all deepfakes the same would be constitutionally suspect at best.”

House TikTok bill faces a minefield in the Senate (By Drew Harwell, Cristiano Lima-Strong, Ellen Nakashima and Jacob Bogage)

How Donald Trump switched to defending TikTok (By Josh Dawsey and Jeff Stein)

How TikTok was blindsided by U.S. bill that could ban it (The Wall Street Journal)

World’s first major act to regulate AI passed by European lawmakers (CNBC)

Google restricts AI chatbot Gemini from answering queries on global elections (Reuters)

Your kid may already be watching AI-generated videos on YouTube (Wired)

Amazon’s hidden chatbot recommends Nazi books and lies about Amazon working conditions (404 Media)

America loves Temu. Will our adoration last for this shopping app? (By Shira Ovide)

• ITIF holds an event, “Transatlantic Innovation and Competitiveness: Europe’s New Agenda vs. the CHIPS and Science Act,” today at 1 p.m.
• The Brookings Institution holds an event, “The dangers posed by AI and disinformation during elections,” today at 2 p.m.
• Public Knowledge holds an event at SXSW, “AI in the Public Interest: Credit, Consent, and Compensation,” today at 3:30 p.m.
• The Federal Communications Commission holds its March open meeting on cybersecurity labels for smart products; cable and satellite TV pricing; and how to define “high-speed” internet on Thursday at 10:30 a.m.

That’s all for today — thank you so much for joining us. Make sure to tell others to subscribe to The Technology202 here. Get in touch with Cristiano (via email or social media) and Will (via email or social media) for tips, feedback or greetings!