Japan Study Finds Over 43,000 AI-Generated Celebrity Images in Two Months as Regulators Begin Review

Over 43,000 AI-generated images and videos that used celebrities' likenesses without consent were created over a two-month period in 2025, according to a study by the Japan Publicity Right Protection Organization. The nonprofit released the findings last month. It documented 43,483 cases of suspected copyright infringement due to AI during the two months from June 2025.

The content drew approximately 335 million views on social media platforms. Financial losses for celebrities and artists were estimated at ¥2 billion to ¥4.5 billion. The calculation used licensing fees for likeness or voice use and the advertising value of the view counts.

The group noted that actual losses could be larger because the study covered only cases it could identify. The survey of 174 entertainment companies showed that 28 percent were fully or somewhat aware of the full scope of damages. Only 1.1 percent said they had guidelines for handling violations of artists' likenesses.

Fifty-two percent said they were currently considering options, while many reported it was impossible to track all illegal uses. The Justice Ministry set up an expert panel earlier this year to discuss possible legal actions against AI-generated content and plans to compile guidelines for the entertainment industry.

Since last December the Japan Fair Trade Commission has been investigating how news content is used without permission by AI-powered search engines.