SDSignal Desk

Schiff calls the White House AI pledge meaningless and floats an ‘FDA for AI’

· Updated · Signal Desk newsroom

Analysis of reporting by The Verge. How we work

Image: The Verge

On The Verge’s Decoder, Sen. Adam Schiff called AI CEOs’ voluntary White House safety pledge meaningless and backed a regulator with real discretion, plus bills on training-data transparency, data-center costs and a narrow antitrust exemption, while acknowledging the post-Loper Bright legal hurdles.

Key takeaways

  • Schiff agreed the non-binding White House AI pledge was meaningless.
  • He floated an “FDA for artificial intelligence” on top of a statutory framework.
  • He described bipartisan training-data transparency work with Sen. Curtis.
  • He backs a data-center bill making operators pay their own costs, and a narrow antitrust exemption against foreign distillation.
  • He said a Democratic Congress would hold oversight hearings on agent incidents.

What Schiff proposed

In a Decoder interview with The Verge’s Nilay Patel, California Sen. Adam Schiff, who sits on Judiciary subcommittees covering intellectual property, privacy and technology, and antitrust, set out how he thinks AI should be governed. The backdrop is the non-binding pact AI CEOs signed at the White House promising to evaluate their own systems (the accord). Asked whether it was toothless, Schiff agreed it was meaningless, and noted that executives signing it say elsewhere that they want actual regulation.

He wants a regulatory body with discretion and expertise, floating an FDA for artificial intelligence, built on a statutory framework because Congress legislates too slowly to keep pace. He described bipartisan work on training-data transparency with Sen. Curtis, a data-center bill he says would force operators to pay their own costs rather than pass them to consumers, and support for a narrow antitrust exemption so labs can coordinate against foreign distillation efforts. With midterms a month away, he said a Democratic Congress would bring vigorous oversight, including asking industry leaders to explain incidents in which their agents got away from them.

The Loper Bright problem

Patel pressed on the hard part. The Supreme Court’s Loper Bright decision ended Chevron deference, so courts no longer defer to agency experts when a statute is vague, and a well-funded lab could challenge a fast-moving AI rule and stall it for years. Schiff’s answer was to write the enabling law as specifically as possible. That is candid and also the bind: the technology moves faster than specific statutes can, while vague statutes now invite litigation. The interview did not resolve that tension.

The narrower proposals are more immediately workable. Requiring companies to retain and disclose copyrighted training material would create an evidence base before fair-use fights. Making data centers bear their own grid costs addresses a real local grievance. A tightly scoped antitrust exemption for security coordination, with guardrails against pricing collusion, mirrors cyber information-sharing laws Schiff says he helped pass a decade ago.

Section 230 and the risk of stalling

Patel argued that when a platform’s own AI generates content, it is plausibly the company’s speech rather than a user’s, and outside Section 230’s shield. Schiff agreed with the analysis but warned that companies can afford to slow litigation. If that legal view takes hold, it could change AI company behavior more than any pledge, though liability fears could also chill defensive tools that are as capable on offense as defense. Schiff compared AI to the failure to address social media harms, arguing the First Amendment was used against rules that were not really speech issues, and cited Flock license-plate cameras and AI surveillance as bipartisan privacy concerns.

The risk is that AI policy is absorbed into partisan fights and bipartisan bills stall as social media reform did, a record Schiff himself called terrible. If that repeats, courts and companies set the rules by default. More in our regulation coverage, future of work coverage and AI coverage.

Who feels it

Frontier AI labs
Possible oversight hearings on agent incidents, training data and liability.
Creators and rights holders
Training-data disclosure would support fair-use fights and licensing.
Communities near data centers
Bills could shift who pays for grid buildout.
Regulators
Any new agency must be designed for a post-Chevron legal landscape.

What to watch

  1. Whether the Schiff–Curtis effort becomes a filed bill with Republican cosponsors.
  2. Midterm results and early AI oversight hearings.
  3. Court tests of Section 230 for AI-generated content.
  4. A concrete standalone AI agency proposal.

FAQ

What did Schiff say about the White House pledge?

He agreed it was meaningless.

What regulator does he want?

A body with discretion and expertise, which he likened to an FDA for AI.

Why does Loper Bright matter?

Courts no longer defer to agencies on vague statutes, so AI rules could be tied up in litigation.

What bipartisan work did he mention?

Training-data transparency with Sen. Curtis.

Read the original

Continue at the source.

The Verge

More on AI

All AI coverage →