New Rules Are Closing In on AI Influencers, and the Creators Behind Them Are Worried

Key points

• The EU's AI Act, the first comprehensive law governing artificial intelligence, requires AI-generated promotional content to be labelled starting 2 August 2025.
• The virtual influencer market was worth an estimated $14.5 billion in 2025 and is projected to reach $110.4 billion by 2033.
• TikTok announced a detection system in mid-July 2025 to catch AI-generated spam it says crowds out real creators.
• 60 percent of creators admitted in 2025 that their AI-assisted content had broken industry rules through mislabelling.
• One EU digital policy adviser warns creators could face overlapping laws that simultaneously declare the same content both lawful and unlawful.

Clarissa Mansbridge spent almost twenty years behind the scenes in entertainment, building brands for pop acts and crisis-managing celebrities. Then, three years ago, her son was born without a heartbeat at 33 weeks. Doctors brought him back. The months that followed, in a neonatal intensive care unit, were terrifying.

When they finally made it home to Brisbane, Mansbridge knew her old career was no longer possible. Her son's recovery needed her close by. So she opened her laptop and started experimenting with AI image-generation tools, software that creates realistic photographs from text descriptions. Out came Mia Metaverse: a glamorous, "humbly wealthy, wholesome" virtual model who exists only as a series of generated images.

Mansbridge, 34, now earns around $6,000 a month from brand deals and building custom AI personas for other companies. Many clients make her sign a confidentiality agreement, so she cannot name them. It is her most profitable year since she started.

But the ground is shifting.

What rules are changing, and why does it matter?

New obligations under the EU Artificial Intelligence Act take effect on 2 August 2025, requiring creators and brands to clearly label any promotional content that has been generated or altered by AI. The law covers images, video, audio and text.

The EU AI Act is the first comprehensive legal framework for AI anywhere in the world. Under Article 50 of the act, anyone deploying an AI system for promotional content must disclose that the content is artificial. There is a narrow carve-out for clearly artistic or satirical work, but it leaves a lot of grey area for virtual influencers who blur the line between entertainment and advertising.

Kai Zenner, digital policy adviser to Axel Voss, a member of the European Parliament who negotiated the Act, told Wired that the law was drafted before policymakers fully understood what they were regulating. "ChatGPT had only just come out," he said. More rules are coming later in 2025 and 2026, requiring creators to disclose not just that their content is AI-generated, but where their funding comes from and what their account is trying to achieve.

Zenner's bigger concern is the collision of overlapping regulations: the AI Act, copyright law, consumer protection rules, and the proposed EU Digital Fairness Act. "People running these accounts could find one law says their content is lawful, while another law says it's unlawful," he said.

Are the platforms making things harder too?

Yes, and quickly. In mid-July 2025, TikTok announced an automated detection system aimed at catching AI-generated spam. Pinterest and Meta have introduced their own labelling and detection tools across Facebook, Instagram and Threads.

Social media consultant Matt Navarra put it plainly: the platforms are "swinging at AI content with a sledgehammer when what they need is a scalpel." Human creators who use AI for small tasks, like correcting a typo or brightening a photo, risk being caught by blunt automated systems designed for something far worse.

Lionel Metge, a music producer in Lyon who created a virtual reggaeton artist named Zairo, has already watched detection tools flag his content almost the moment he posts it. His account has 23,800 followers. He accepts the rules are necessary to stop the public being misled, but worries the algorithms will overshoot.

Not everyone is anxious. Rob Haywood, head of safety at Fanvue, a creator-monetisation platform, says many creators are choosing to lean into transparency rather than fight it. Some are building content that only makes sense coming from an AI persona, making the artificial nature of the account the whole point.

Mansbridge is thinking the same way. She is building a portfolio of AI characters she hopes to cross-promote and eventually run with minimal day-to-day input. Her goal is a financial cushion for her son, whatever he needs.

"I want to create a legacy for my son," she said. Making AI content that is openly, unapologetically artificial may turn out to be the safest bet of all.

What should ordinary people watch for?

If you follow accounts on social media, a label reading "AI-generated" or "AI info" does not mean the content is scam or spam. It may simply be a creative persona built by a real person. The disclosure requirement is there so you know you are not watching a human being.

The real risk for consumers is AI content with no disclosure at all: a virtual spokesperson selling a product without any indication they do not exist. If you see a polished influencer promoting a financial product, a health supplement, or an investment scheme and there is no label, no bio detail, and no sign of a real person behind the account, treat it with extra caution before you buy.