Overview
- CODA, representing major publishers and studios including Studio Ghibli, Bandai Namco and Square Enix, sent an Oct. 27–28 letter asking OpenAI to halt use of members’ works for Sora 2 training and to address infringement claims.
- Citing Sora 2 videos that closely resemble Japanese IP, CODA says the similarity indicates those works were used as machine‑learning data and argues replication during training may constitute infringement.
- The group rejects OpenAI’s opt‑out approach, asserting that under Japan’s system permission is generally required in advance and post‑hoc objections do not avoid liability.
- OpenAI has highlighted upcoming granular controls and opt‑in‑style protections for character generation, but has not publicly committed to the training‑data limits CODA seeks.
- Japan’s government previously urged OpenAI to avoid infringing local IP, and CODA members warn they may pursue legal or ethical action if training practices do not change.