Generated by Codex with GPT-5

What happened

Techmeme surfaced Emma Roth’s May 12, 2026 article for The Verge, and the original article is George Clooney, Tom Hanks, and Meryl Streep back new ‘Human Consent Standard’ for AI licensing. The Techmeme item framed the announcement as a new AI licensing effort from RSL Media, backed by high-profile creative figures and entertainment organizations.

The Human Consent Standard is meant to let people declare how AI systems may use their likenesses, voices, creative works, characters, marks, and designs. The basic options are broad permission, conditional permission, or restriction. In practical terms, it tries to give people and rights holders a machine-readable way to say whether an AI company can train on, generate from, or otherwise exploit an identity or work.

The important technical detail is that this is not just another website crawling rule. RSL Media is building on the Really Simple Licensing standard, which already lets publishers attach licensing and compensation terms to web content. But the Human Consent Standard aims at the underlying person, work, character, or brand, even when it appears in many places across the internet. That distinction matters for likeness rights: an actor’s face, a musician’s voice, or a fictional character is not confined to one URL.

The system is also tied to a registry expected to launch in June. People and rights holders would verify identity, publish their consent preferences, and have those declarations translated into signals that AI systems can check. The Verge reports support from figures including George Clooney, Viola Davis, Tom Hanks, Kristen Stewart, Steven Soderbergh, and Meryl Streep, plus organizations such as Creative Artists Agency and the Music Artists Coalition.

Why it matters

The announcement is interesting because it moves the AI rights debate from lawsuits and opt-out forms toward infrastructure.

Most disputes over AI training data and synthetic media have been handled after the fact: a model is trained, a voice clone appears, an image generator mimics a style, or a chatbot emits material that looks too close to a protected work. Then the fight shifts to takedown requests, contract arguments, copyright claims, right-of-publicity law, platform policies, or public pressure. A machine-readable consent system tries to move some of that governance earlier in the pipeline.

If it works, AI developers would have a more standardized input for data ingestion, generation filters, licensing workflows, and audit trails. Instead of treating identity and creative rights as a pile of one-off legal exceptions, they could check a registry and decide whether a given use is allowed, restricted, or requires negotiation. That would not settle every legal question, but it could make compliance less ad hoc.

The standard also reflects a widening of the AI licensing conversation. Publisher-focused RSL is mostly about websites and content economics: whether AI crawlers can use articles, code, images, and archives, and on what terms. The Human Consent Standard is about personal and creative control. It tries to cover the cases where the asset being copied is not only a file, but a person’s image, voice, name, performance, or recognizable creation.

That is why celebrity backing matters without being the whole story. Famous actors and musicians have the money and leverage to fight unauthorized uses, and some are already pursuing trademarks or other rights strategies. The more important claim from RSL Media is that ordinary people should have a place to publish the same kind of preference signal. Deepfake and voice-cloning risk is not limited to public figures, and neither are disputes over whether AI systems can imitate a person’s work.

The hard part

The weak point is adoption.

A consent standard only changes behavior if AI companies, platforms, crawlers, and downstream application developers actually check it and honor it. Robots.txt became useful because many crawlers agreed to treat it as a norm, not because it was technically impossible to ignore. RSL and the Human Consent Standard face the same coordination problem, with higher stakes and more complicated incentives.

There is also a verification problem. A registry must prove that the person setting terms really controls the relevant identity, work, character, or mark. That is straightforward for some public figures and brands, but harder for disputed works, collaborative creations, fan works, aliases, estates, and ordinary people whose likeness is scattered across third-party platforms. The more valuable the registry becomes, the more it will attract impersonation, conflicting claims, and pressure to resolve rights disputes that may not have clean technical answers.

Still, the direction is notable. AI companies increasingly need a way to show that their training and generation systems are not simply ignoring human consent. Creators increasingly need something more scalable than lawsuits. A registry-backed, machine-readable standard is not enough by itself, but it is the kind of boring infrastructure that could make future licensing markets possible.

Takeaway

The Human Consent Standard is best understood as a bet that AI rights will need protocol-level machinery, not just case-by-case legal fights.

Its immediate effect may be limited until major AI labs and distribution platforms commit to honoring it. But the idea points to where the market is going: consent, licensing, attribution, and compensation need to become inputs that software systems can process before models train or generate output. If RSL Media can make the registry trustworthy and get broad adoption, the standard could become a useful layer between creators and AI developers. If not, it may still become evidence of a changing norm: people expect machine-readable ways to say no, yes, or pay me before AI systems use their identity and work.