AI copyright report sparks new fight

The big picture: The use of copyrighted material to train AI models has been controversial since the dawn of ChatGPT, with some publishers and content creators striking deals, some suing and others waiting to see how things shake out.

• Publishers and creators say the firms' unauthorized and uncompensated use of massive quantities of copyright text, video and audio is simply a giant rip-off.
• The firms — arguing that their AI models are just learning from this material the way a student might — say they're protected by copyright's "fair use" principle.

Driving the news: The Copyright Office's "pre-publication" version of its 108-page report argues for a balanced approach that recognizes the contributions of both tech firms and content creators.

• "American leadership in the AI space would best be furthered by supporting both of these world-class industries that contribute so much to our economic and cultural advancement," the agency writes.

The report concludes that while some generative AI probably does constitute a "transformative" use, the mass scraping of all data for commercial use probably does not qualify as fair use.

• "The extent to which they are fair, however, will depend on what works were used, from what source, for what purpose, and with what controls on the outputs — all of which can affect the market," the report states.
• "Making commercial use of vast troves of copyrighted works to produce expressive content that competes with them in existing markets, especially where this is accomplished through illegal access, goes beyond established fair use boundaries."

The report encourages the U.S. government to help foster the nascent market for licensed content for AI training.

Yes, but: The Trump Administration, without directly commenting on the report, almost immediately fired Shira Perlmutter, the head of the Copyright Office.

• Perlmutter's firing — like the dismissal of the librarian of Congress that preceded it — broke from precedent, since both the Copyright Office and the library are under Congressional oversight.

Zoom in: Democrats argue Perlmutter's removal is illegal.

• "Donald Trump's termination of Register of Copyrights, Shira Perlmutter, is a brazen, unprecedented power grab with no legal basis," Rep. Joe Morelle (D-N.Y.) said in a statement. "It is surely no coincidence he acted less than a day after she refused to rubber-stamp Elon Musk's efforts to mine troves of copyrighted works to train AI models."
• "The President lacks the legal authority to terminate the position unilaterally," Sens. Charles Schumer (D-N.Y.) and Adam Schiff (D-Calif.) said in a statement. "Congress purposefully insulated this role and the U.S. Copyright Office from politics and did so on a bipartisan basis."

Between the lines: Many AI firms have trained their systems on almost anything found on the open web, what they like to call "publicly available" information — a category so broad that in some cases it includes known archives of pirated content.

• OpenAI, among other tech firms, used recent comments on U.S. AI strategy to call for an enshrined right to train on copyrighted data in order for the U.S. to keep its lead over China, which has a long record of violating copyright restrictions.

• "Whoever ends up winning ends up building the AI rails for the world," Chief Global Affairs office Chris Lehane told Axios during an Axios event in March.

The other side: Content creators spanning music, film, visual art, publishing and journalism have sued, and a number of actions are working their way through the courts.

• Representatives from creative industries also submitted comments of their own as part of the White House call for input on national AI strategy,
• The News/Media Alliance, which represents newspaper and magazine publishers, says it is not looking to stifle AI development, but argues for consent and compensation.
• "Publishers should not be forced to subsidize the development of AI models and commercial products without a fair return for their own investments, no more than cloud providers would be expected to bear the costs of compute without payment for their input," the organization said in its filing.

My thought bubble: OpenAI's powerful new image generator highlights the risks of giving AI companies total freedom to digest the entire universe of intellectual property.

• ChatGPT now enables users to turn any portrait into a "Simpsons" character, a "Sesame Street" Muppet, a Studio Ghibli animation or creatures from countless other fictional universes. The creators of those worlds and styles have no say and get no reward.

Disclosure: Axios and OpenAI have a licensing and technology agreement that allows OpenAI to access part of Axios' story archives while helping fund the launch of Axios into four local cities and providing some AI tools. Axios has editorial independence.