The federal government recently sought input on what should be included in a federal “AI Action Plan,” and the responses of two companies drew a series of stark headlines: “OpenAI and Google ask for a government exemption to train their AI models on copyrighted material,” “OpenAI and Google ask the government to let them train AI on content they don’t own” and “Google joins OpenAI in pushing feds to codify AI training as fair use.” Those were from outlets that specialize in technology; general interest news sites were no better. These headlines were so alarming, they sent Hollywood A-listers rushing to Google Docs to fire off an open letter saying the AI companies were “arguing for a special government exemption” for “so-called fair use” and had to be stopped. The vultures at Alden Global Capital, whose entire business model is just destroying local news, fired off a similarly misinformed screed saying the feds shouldn’t give AI companies any new rights to use copyrighted works. 

Here’s the thing, though: that’s not what Google’s comments or OpenAI’s comments say. Both companies explain that their AI training activities are already legal under existing fair use law. They don’t ask for any changes to copyright, because there’s no need to expand or codify anything, no need for the government to grant permissions or new exemptions, no need to “weaken copyright.” As they have told the courts (and leading scholars agree), copyright is not a barrier to AI training because copyright includes fair use. Indeed, they point out that fair use gives the US a strategic advantage over jurisdictions with less flexible copyright laws. Copyright might become a barrier to AI development if courts swerve in the wrong direction, but there’s no strong reason to think they’ll do that, yet. (As I explained in a previous Note, the district court opinion in Thomson Reuters v. Ross was a dead end, and it’s apparently headed up on appeal.)

Fair use is flexible and open-ended by design. Congress codified fair use in the 1976 Copyright Act and said explicitly that fair use empowers courts to continue applying broad principles flexibly in response to new situations. When artists and technologists invoke fair use, they aren’t asking to change copyright law. To the contrary, they are exercising a First Amendment right deeply embedded in existing law. Fair use preserves the balance of copyright law when technology threatens to turn copyright into a tool for censorship and anticompetitive behavior.

The truly radical change would be allowing copyright holders to leverage what’s supposed to be a limited monopoly in a particular work or portfolio of works into a right to control and monopolize entire fields of technology. That kind of control is something Congress did not foresee or intend when it granted copyright holders the right to make and sell copies. Fair use is designed to prevent the abuse of copyright to stifle competition and free expression. Protecting AI training is a classic example of fair use in action.

I know some reporters receive this newsletter. If you’re interested, I’ve included below a few headlines and subheds you could use if you write about the content industry’s comments on AI policy—I don’t even need attribution as the author:

Content industries demand new exclusive right to block statistical analysis – Ownership of facts would help boost profits, they say.

Publishers ask the government for power to block “unlicensed” writing. –Without this power, they warn Americans could be overwhelmed by polished, professional writing by authors not under contract to publishers.

Recording industry seeks power to tax independent music creation – Since nothing comes from nothing, everyone must pay.