Amazon Ran an AI Dubbing Trial on Real Subscribers Without Telling the Studios

What happened

In November 2025, Amazon Prime Video began serving AI-generated English audio tracks on a handful of anime titles, including Banana Fish, No Game No Life: Zero, and Vinland Saga. The tracks appeared in the audio menu labeled "English [AI beta]," a small tag that was, in effect, the only notice subscribers received before finding that the voice performances they were hearing had been generated by a machine.

Clips circulated quickly. The delivery was flat, the timing was off, and the AI audio frequently diverged from the subtitles on screen. Voice actors and fans alike recognized the output as substandard. Daman Mills, known for his work on Dragon Ball, called the move a massive insult to performers. The National Association of Voice Actors issued a formal condemnation. Some subscribers cancelled their Prime memberships outright, naming Crunchyroll as the platform they would use instead.

The backlash widened when several anime studios confirmed they had not been consulted or given approval before Amazon applied AI dubbing to their content. That fact separated this episode from an ordinary quality dispute. It was not simply that the dubs were poor. Amazon had applied a novel production method to titles it did not originate, without the knowledge of the creators who held rights over the work being altered. The studios said they were looking into the situation with Amazon, suggesting they learned about the deployment the same way everyone else did.

Amazon pulled the AI English audio tracks from Banana Fish and No Game No Life: Zero within days, though Spanish AI tracks for some titles remained. The rollout appears to have been driven by a push to reduce localization costs and accelerate content delivery across languages. The "AI beta" label was Amazon's primary disclosure mechanism, and it offered no explanation to labor groups, unions, or studios ahead of deployment. Workers in anime dubbing, an industry sector already operating on thin margins, were given no notice that their work was being replaced on live titles.

The gap the incident exposes is not primarily about audio quality standards, though those matter. It is about the absence of any prior record showing that the studios consented, that affected workers were notified, and that the content had been authorized for AI alteration before it went live. A provable record of what a system did, when it was deployed, and who approved each step would have made those questions answerable before the backlash rather than only afterward, in the form of pulled tracks and public statements.

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.