Skip to content
Submit incident
AII-1002-01879
Well documentedThree or more independent publishers report the core facts.Supporting sources are pending verification.

Crunchyroll Uses ChatGPT for Anime Subtitles With Unreviewed Errors

Crunchyroll aired anime with ChatGPT-generated subtitles containing unreviewed errors, typos, and explicit references to the AI system. The incident contradicted the company's earlier statement that it was not considering AI in its creative production process.

2 July 2025 Β· Crunchyroll Β· Reliability and hallucination
2 July 2025
AII-1002-01879
Quick facts
Category
Reliability and hallucination
Country
United States
Date
2 July 2025
Drafted by
Raidu Incident Detection

What happened

Crunchyroll's subtitles for Studio Gokumi's Necronomico and the Cosmic Horror Show contained typos, grammatical errors, and phrases like "Is gameorver. if you fall, you are out," indicating the text was generated by ChatGPT without human review or editing.

Some subtitles explicitly began with "ChatGPT said," confirming the use of the AI system. Crunchyroll did not respond to questions about how the subtitles were produced or what steps were taken to prevent such errors.

Reported impact

Viewers were surprised and critical of the low quality subtitles and the lack of human oversight in the localization process. Fans expressed concerns about AI replacing creative roles without proper review, particularly in localization where context and tone are crucial.

Index analysis

The incident contradicts Crunchyroll president Rahul Purini's statement to Forbes in April 2025 that the company was "not considering AI in the creative process" to maintain authenticity and would not use AI in ways impacting voice actors.

Governance controls

Crunchyroll stated it was actively looking into ways to use AI for discoverability, recommendations, and personalization, but did not disclose plans to use AI for subtitle generation.

Classification

Category
Reliability and hallucination
Industry
Media, social platforms and marketing
Country
United States
Harm status
Actual harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Reputational
Harmed
Consumers
Failure modes
Hallucination
Consequences
None recorded
AI systems
ChatGPT

Sources and evidence

Sources not yet published

This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.

How sources are verifiedSuggest a source for this incident