FTC takes aim at generative AI in rare copyright office filing

AI startups thought their biggest legal headache was fighting over Fair Use in front of a copyright judge. But the Federal Trade Commission just opened up an entirely new flank.

In a rare and aggressive filing submitted to the U.S. Copyright Office, which has been studying AI and collecting public comments since August 2023, the FTC made a bold claim: even if training AI on copyrighted content passes muster under copyright law, it could still be illegal under consumer protection statutes.

Section 5 enters the chat

The FTC's core argument leans heavily on Section 5 of the FTC Act, which bars unfair or deceptive business practices as well as unfair competition. The agency warned that generative models frequently train on pirated materials scraped without consent, which could constitute an unfair method of competition on its own.

Deception is the other red flag. The FTC pointed out that AI tools can easily mimic an artist's face, voice, or performance without permission. When consumers think a piece of media was created by a human artist, but it was actually generated by software, consumer harm has already occurred.

The most controversial part of the filing centers on the Fair Use doctrine, the legal defense currently relied upon by tech giants like OpenAI and Anthropic in ongoing litigation. According to the FTC, conduct that is technically permissible under copyright law does not grant a free pass from consumer protection laws. In the agency's words, there is no AI exemption from the laws on the books.

Tech pushes back against Lina Khan

The tech industry was quick to call foul. In an interview with VentureBeat, Adam Kovacevich, CEO of the tech coalition Chamber of Progress, accused the FTC of massive regulatory overreach.

Kovacevich argued that the FTC's founding charter contains zero references to copyright, an area traditionally resolved strictly by the courts. He characterized the move as an effort by FTC Chair Lina Khan to unilaterally expand the agency's enforcement mandate.

He also challenged the FTC's antitrust logic. According to Kovacevich, copyright itself functions as a monopoly right, whereas Fair Use has always acted as an original antitrust safeguard. If every AI model requires paid licensing for every piece of training data, only deep-pocketed incumbents will survive, effectively wiping out early-stage startups that lack the capital to cut massive licensing checks.

Why it matters

We think the FTC's maneuver fundamentally shifts the risk profile for anyone building or deploying generative AI models.

Up until now, the AI industry's entire defense strategy rested on winning the Fair Use argument in federal court. If the FTC sticks to this filing, that strategy is no longer a silver bullet. Even if a court rules that web scraping for model weights is fair use, a regulator can still hit a developer with unfair competition or consumer deception charges if the data was sourced from pirate hubs or if outputs mimic existing creators too closely.

There is also a glaring paradox in the FTC's antitrust posture. While the commission wants to protect creative workers and keep powerful platforms in check, weaponizing unfair competition claims against data scraping could easily backfire. Big tech companies can write licensing checks to publishers; cash-strapped startups cannot. By eroding Fair Use, the FTC might accidentally entrench the very Big Tech monopoly it claims to fight.

For builders and enterprise users, the immediate takeaway is clear: provenance and transparency matter more than ever. Disclosing AI involvement and auditing training sources isn't just about avoiding a DMCA notice anymore; it's about staying off the FTC's deceptive practice radar.