Silicon Valley’s new gold rush is built on stolen work

Lawmakers and the public have long grown tired of Silicon Valley’s “move fast and break things” ethos. But Big Tech hasn’t learned its lesson.

In the mad rush to roll out artificial intelligence, companies have cast aside basic corporate responsibility. Meta’s chatbots allowed children to engage in sexual conversations. OpenAI’s ChatGPT helped a child plan and commit suicide. And mounting evidence shows AI chatbots harming mental health, particularly for children.

OpenAI has created a copyright infringement factory.

The latest example of Silicon Valley’s negligence? Mass copyright infringement.

The wild digital west

When AI products like ChatGPT were new, copyright infringement — while potentially common in AI training — was more opaque. The early versions of today’s large language models weren’t really capable of producing realistic copies of original copyrighted works that could compete with the originals. Moreover, training data was hidden from public view.

But new releases, particularly OpenAI’s Sora 2, have smashed that paradigm. This past week, X and other social media platforms have been flooded with Sora 2 users’ reproductions of TV shows like “Family Guy,” “South Park,” “SpongeBob,” and more. While these clips are not yet perfect replicas, they are extremely close, and it is likely a matter of years — if not months — before users can produce copyright-infringing content that is essentially indistinguishable from the originals.

Flipping the script

OpenAI initially put the onus on copyright holders to “opt out” of their creations from being used in Sora 2. But this simply is not how copyright law works. In fact, it violates the express purpose of copyright law: to ensure that your work stays yours. Others have to ask you for permission to opt in to using your work; reversing that would render copyright law toothless.

After the understandable outcry from copyright holders and actors’ guilds, OpenAI changed its copyright policies, switching from the opt-out model to opt-in. Rights holders now choose whether to have their creations included in Sora 2.

CEO Sam Altman seemed to imply that he was surprised at the backlash. “I think the theory of what it was going to feel like to people, and then actually seeing the thing, people had different responses,” he told the Verge. “It felt more different to images than people expected.”

This dubious claim is not an excuse to violate copyright laws wantonly. In waiting days to change course, Altman has also opened a Pandora’s box: the idea that AI can create entire seasons of your favorite shows. Altman knows that users will want it and may hope that the pressure will force copyright holders to take a deal that doesn’t fully represent the value of their intellectual property.

Altman has said as much. “We are going to try sharing some of this revenue with rights holders who want their characters generated by users,” he wrote recently. His questionable usage of “try” notwithstanding, the game he is playing is clear: Audiences are going to want this. You either give us permission — after which we will “try” to compensate you — or people will likely use AI to violate your copyright anyway.

RELATED: Does anyone think we’re up to the task of controlling AI?

Mininyx Doodle via iStock/Getty Images

Congress must act

None of this will eliminate AI’s chilling ability to generate lifelike images of people. While the Trump administration cracked down on some of the worst offenders with the Take It Down Act — which bans involuntarily created AI porn — you can still easily create false images of people doing all sorts of things.

For now, the predominant use case may be “funny” videos of SpongeBob fleeing from police. But it would be a simple task, for example, to replace SpongeBob with realistic videos of innocent people committing crimes.

OpenAI and other AI companies have, in essence, created a copyright infringement factory that puts the onus on copyright holders to vindicate their rights, while signaling that they may leverage a rabid user base if copyright holders don’t get in line. This isn’t right, and it shouldn’t be how the law works. Given how flippantly OpenAI and other Big Tech companies have treated the law, Congress needs to consider legal clarifications that force these firms to change their behavior.