Schiff Urges Faster AI Legislation and Backs a Dedicated Regulator

Congress cannot spend years debating AI rules as it did social media, Sen. Adam Schiff argued in an October 5, 2026, interview with The Verge’s Nilay Patel. The California Democrat backed a dedicated AI regulator and called for companies to preserve records of copyrighted training material before lawmakers tackle which uses qualify as fair use.

A regulator, not just a pledge

The interview addressed a nonbinding White House pledge by technology executives to evaluate their AI systems. Schiff criticized the contrast between leaders signing voluntary commitments and, in his account, calling for regulation away from the president. He rejected reliance on companies promising to do the right thing.

A new agency would also face a legal constraint highlighted by Patel. The Supreme Court’s Loper Bright decision overturned a precedent under which courts deferred to agency experts when laws were vague. Patel raised the prospect that rules from a new AI regulator could quickly become entangled in litigation.

Why social media stalled

Patel challenged Schiff’s comparison with social media. He argued that efforts to control disinformation, bullying and harmful content repeatedly ran into First Amendment protections. In that reading, the obstacle was not simply congressional delay: proposals aimed at harmful posts became attempts to regulate speech, while privacy legislation remained caught alongside them.

Schiff disagreed. He said companies invoked the First Amendment against measures that did not necessarily concern speech. He instead identified their broad immunity under Section 230 as a threshold question for regulation. His warning about AI was more urgent: he argued that its potential harms were greater and Congress could not repeat its slow response.

It can’t wait. We certainly can’t take the kind of years we’ve taken to try to even wrap ourselves around social media.

Sen. Adam Schiff, speaking to The Verge’s Decoder

Training records before copyright judgments

On copyright, Schiff offered a sequence rather than a blanket answer about AI training. He said he and Sen. John Curtis were working to require companies to retain information about protected material used in their models. Establishing what went into training would come first; debating fair use would follow.

That position has an earlier legislative counterpart. On February 10, Schiff and Curtis introduced the Copyright Labeling and Ethical AI Reporting, or CLEAR, Act. According to Schiff’s official announcement, the proposal would turn training-data transparency into disclosure duties, rather than leave the information solely in company records.

• Before a new generative AI model’s public release, companies would submit a notice to the Register of Copyrights summarizing copyrighted works used in its training dataset.
• The Copyright Office would maintain a publicly accessible online database of those notices.
• The proposal would also apply retroactively to models already available to consumers and impose civil penalties for failures to submit required disclosures.