Voice Actors Were Ready to Perform a GPT-4 Script. Rights Holders Were Not.

What happened

In March 2024, a Tokyo production company called Lol announced an unusual theatrical event: professional voice actors would perform a romantic drama script generated entirely by AI tools, including GPT-4. The organizers billed the format as a feature, not a flaw. The stated concept was that actors would deliver the material "regardless of any unnatural content or plot points," a kind of live experiment in AI-generated performance. Tickets were sold. A premiere date was set.

Lol did not write the script themselves. The company subcontracted the work to a creator who used multiple AI tools to produce the drama. That creator's process was not obscure; GPT-4's involvement was acknowledged from the start. What was not addressed at the time of the announcement was where those tools learned to write like drama writers in the first place.

Critics made that gap explicit. They argued that the AI tools used to generate the script had been trained on copyrighted works, including scripts and other literary material, without the permission of the original creators or rights holders. The complaint was not merely theoretical. Lol was accused of building a commercial ticketed event on the output of systems that had ingested other writers' work without compensation or consent. That framing made the production not just an experiment in AI performance but, critics argued, an exploitation of creative labor dressed up as a novelty act.

The backlash was fast and substantial. Four days before the planned premiere, Lol cancelled the event. In their statement, the company acknowledged what they called a failure to explain their intentions sufficiently, a formulation that treated the problem as one of communication rather than conduct. No legal action followed, which means the underlying copyright questions the incident raised were never formally adjudicated.

That absence is the gap this incident points to. The commercial use of AI-generated content, built on models trained on copyrighted source material, currently requires no disclosure to the rights holders whose work shaped those models. Lol knew what tools they used; what they did not have to prove, and had no system to prove, was whether those tools held any license for what they learned from. A provable record of what a system was trained on, and what permissions governed that training, would have forced that question into the open before the tickets went on sale, not after public pressure made cancellation the only available exit.

Reported impact

Affected parties

Not publicly disclosed

Harm type

Not publicly disclosed

Scale

Not publicly disclosed

Financial impact

Not publicly disclosed

Regulatory action

Not publicly disclosed

Classification

Organization

Not publicly disclosed

AI system

Not publicly disclosed

Industry

Not publicly disclosed

Country

Not publicly disclosed

Provider

Not publicly disclosed

Incident type

Not publicly disclosed

Relevant governance controls

Governance control mapping is not available for this record.

• No controls mappedNot publicly disclosed

Control mapping is analytical. It does not state that any control would have prevented the incident.

Sources and evidence

This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.