To Their Credit

Games

Six European game-worker unions demand curbs on generative AI

Their manifesto calls for worker control, clear labels and consent and payment before creative work enters AI training datasets.

Ledger: ResistedAI is held back

Six European game-worker unions have issued a joint manifesto urging developers, studios and players to reject generative AI in game production, while calling for a regulatory ban on adding a person’s work to AI datasets without consent and fair payment. The document argues that the technology is worsening conditions, eliminating jobs and diminishing the games made with it, warning that the effects will deepen if its use goes unchecked.

The manifesto, “On Generative AI Usage in the Games Industry,” was led by the IWGB Game Workers’ Branch in Great Britain. It was written with Game Workers Unite Ireland, Spain’s CSVI and the FIOM-CGIL Milan Work Council, while Germany’s GDRT and France’s STJV also took part in the Manchester summit behind the initiative.

Its authors define generative AI as systems that create text, images, video or music from limited inputs, usually prompts. They say workers are often compelled to use the tools either by explicit instruction or because performance expectations cannot otherwise be met.

The unions contend that the technology is being used to automate work and reduce headcount, including roles traditionally assigned to junior staff. That, they say, narrows the route into games for newcomers. The manifesto also says some employees have been made to train models that subsequently replaced them.

It argues that supervising and correcting AI output can make work more tedious and less fulfilling, while reported short-term productivity gains fade. Workers have reported reduced focus, cognitive fatigue, burnout and impaired decision-making, according to the document.

The text also says employers can use AI systems’ agreeable, human-like interfaces to make unreasonable requests or expand workloads without regard for real-world limits. It warns that companies may shift responsibility for errors to AI platforms, or blame staff for failures arising from mandated use, while detailed human checking can erase the promised efficiency gains.

On intellectual property, the manifesto alleges that most generative-AI models rely on stolen datasets and that employers frequently place staff contributions into internal training datasets without consent or additional compensation. It also raises concerns about discrimination, saying systems can reproduce racism, sexism, homophobia and transphobia, and produce more homogeneous characters and viewpoints.

The unions call on workers to avoid generative AI where possible, explain their objections and organize. “We do not have to accept the use of generative AI to make video games; it is not inevitable,” the manifesto says.

For employers, the proposed standard is that workers should decide when, where and how the technology is used. Studios should ensure nobody is required to use it and prohibit it in codes of conduct, including at the prototyping stage; where use cannot be avoided, it should be clearly labelled. Players are asked to support games made without generative AI and publicly flag uses they identify.

The regulatory demand is more direct: no person’s work should be added to a generative-AI dataset without explicit consent and fair payment. The manifesto groups its objections into five areas: deteriorating conditions, strained worker-employer relations, appropriation of workers’ output, discrimination and hidden costs.

The intervention comes as use and unease rise together. Game Developer’s 2026 State of the Game Industry survey found that 40% of professionals used generative-AI tools in their work, up from 36% a year earlier. Yet 55% said the technology was having a negative effect on the industry, compared with 35% who described its impact as mixed and 8% who saw it as positive.

A separate study based on interviews with 20 Belgian game developers found broad skepticism about generative AI’s usefulness, although some participants saw potential in limited supporting roles, particularly coding and early-stage development. Quality and job-loss concerns were persistent, and developers were more open to tools trained on open-source or voluntarily supplied data than broad conversational systems.

See also

More from Games

All Games news
  1. Mikami and Kim split over AI’s role in game developmentLedger: Disputed
  2. Ubisoft turns its Teammates team into Motherbrain AI subsidiaryLedger: Adopted
  3. France’s CNC withholds funding from AI-only dubbing and subtitlesLedger: Resisted