What started as a dispute over a single AI voice has grown into one of 2026’s biggest fights between the entertainment industry and AI companies: whether a person’s face, voice, and likeness can be recreated by AI without their consent, and what happens when it is. The controversy now has real legal stakes, active congressional attention, and a specific federal bill that could determine how it ends.

How it actually started

The clearest starting point traces back to 2024, when OpenAI asked actress Scarlett Johansson to license her voice for a new ChatGPT voice assistant. She declined. When OpenAI released the assistant anyway with a voice widely described as strikingly similar to hers, called “Sky,” Johansson threatened legal action, arguing OpenAI had effectively used her voice without permission after she’d already said no. OpenAI paused the voice and denied it was modeled on her, but the episode became the reference point every later incident got compared against.

The dispute escalated well beyond voice cloning. In early 2025, an Israeli creative agency used generative AI to simulate the faces of Johansson and roughly a dozen other Jewish celebrities in a video responding to antisemitic remarks, without their involvement or consent, raising a separate question about AI being used to put real people’s likenesses into political statements they never made. Later that year, users of Grok Imagine, the AI image and video tool built into Elon Musk’s X platform, found they could generate photorealistic and sexually explicit content of Johansson and other celebrities on request, pushing the controversy from consent-over-commercial-use into a much more serious harm: non-consensual sexual content generated from a real person’s likeness.

Why this became a policy fight, not just a celebrity dispute

Johansson isn’t the only figure driving this. She’s joined roughly 700 other entertainment industry professionals in the “Stealing Isn’t Innovation” campaign organized by the Human Artistry Campaign, arguing AI companies are training models on copyrighted creative work and generating unauthorized likenesses without any licensing framework at all. A congressional panel has separately taken direct interest in her specific case, inviting her to speak to lawmakers about AI voice and likeness misuse.

What makes this different from most celebrity-versus-tech-company disputes is that current US law genuinely doesn’t have a clear, unified answer. Right-of-publicity law, which governs unauthorized commercial use of someone’s name or image, exists at the state level and varies significantly, there’s no single federal law that directly addresses an AI system generating a realistic video or explicit image of a real person without consent. That gap is exactly what the pending federal legislation is trying to close.

How this actually gets resolved: our prediction

The most concrete path to resolution is the NO FAKES Act, a bipartisan federal bill that would create, for the first time, a nationwide legal right against unauthorized AI-generated digital replicas of a person’s voice or visual likeness. As of mid-2026, the bill has cleared the Senate Judiciary Committee unanimously and moved to the full Senate, a genuinely significant step for federal legislation, but it has not yet passed either chamber of Congress and is not yet law.

Based on the bill’s unanimous committee support and the volume of public attention this issue has attracted through 2026, we expect some version of the NO FAKES Act, or closely related state-level laws filling the gap in states that move first, becomes real, enforceable law within the next two to three years, rather than stalling out entirely. What’s less certain is enforcement in practice: a federal right of publicity gives people like Johansson legal standing to sue, but platforms generating millions of AI images and videos daily will still need to build detection and takedown systems that actually work at scale, and that technical problem doesn’t get solved just because a law passes.

The more likely near-term outcome is a patchwork: individual platforms tightening their own content policies under public pressure, the way X faced backlash over Grok Imagine’s outputs, combined with a slow federal legislative process that eventually produces a real law but takes longer than the controversy itself.

What people get wrong about this controversy

This isn’t only about famous people. Celebrities have the visibility and legal resources to make these cases public, but the same AI tools that generate unauthorized celebrity content can be pointed at private individuals, who have far less recourse and far less attention on their case.

AI companies denying intent doesn’t settle the legal question. Whether a company deliberately trained a model to mimic a specific person, or a model simply produces a similar-sounding voice or face because it was trained on enough public content featuring that person, is a real legal distinction, but it doesn’t change the impact on the person being replicated, which is why the proposed law focuses on the output, not the intent behind training.

There is currently no single federal law covering this. Existing protections vary state by state and were largely written before generative AI existed, which is precisely the gap the NO FAKES Act is designed to close.

Frequently asked questions

How did the Scarlett Johansson AI controversy start?

It began in 2024 when she declined to license her voice to OpenAI for a ChatGPT voice assistant, then OpenAI released a voice, later paused, that many listeners said sounded strikingly similar to hers.

What is the NO FAKES Act?

It’s a bipartisan federal bill that would create a nationwide legal right against unauthorized AI-generated replicas of a person’s voice or likeness. It passed the Senate Judiciary Committee unanimously in mid-2026 but has not yet become law.

Is there currently a federal law against AI deepfakes of real people?

No. Protections currently exist only at the state level through right-of-publicity laws, which vary widely and generally weren’t written with generative AI in mind. The NO FAKES Act would be the first federal law directly addressing this.

What is the “Stealing Isn’t Innovation” campaign?

It’s an advocacy effort from the Human Artistry Campaign, joined by around 700 entertainment industry figures including Johansson, arguing that AI companies are using copyrighted creative work and real people’s likenesses to train and generate content without licensing or consent.

When will this issue actually be resolved?

There’s no fixed timeline. The NO FAKES Act has strong committee-level bipartisan support, which suggests real momentum, but it still needs to pass both chambers of Congress. A realistic expectation is federal legislation within the next two to three years, alongside individual platforms adjusting their own policies sooner under public pressure.

For more on how AI is actually reshaping industries and rights, not the hype-cycle version, see Talmyn’s Tech & AI Evolution desk.