Sony trains 'Protective AI'
Sony has trained a 'Protective AI' on Studio Ghibli material to automatically block infringing generative content — the company says creators will be compensated for use while AI prevents unauthorized outputs. It's a notable case of a studio building guardrails specifically tuned to a single creative corpus. (x.com)
The scoop was first reported by The Nikkei and was carried widely in the West when IGN published a write‑up on March 18, 2026. (ign.com) Sony AI — the Sony Group research arm — has been explicit about publishing research on attribution and protection, including a December 2025 blog post titled “Protecting Creator’s Rights in the Age of AI.” (ai.sony) Reporting that cites Nikkei says Protective AI is being trained to refuse not only explicit “Ghibli‑style” prompts but also more subtle, indirect prompts intended to reproduce the studio’s look. (automaton-media.com) Separate Sony research described in Nikkei coverage and summarized by outlets shows the company has built tools to detect original songs inside AI‑generated tracks and estimate how much each source contributed — a metric Sony says could underpin future revenue‑sharing. (digitaltrends.com) Sony’s moves sit alongside industry work: Sony Music publicly backed the “Stealing Isn’t Innovation” campaign in January 2026, and Sony has collaborated with the SoundPatrol research lab alongside Universal Music to protect artists from AI misuse. (sonymusic.com) (musicweek.com) Sony characterizes Protective AI as a research‑stage effort with no announced product rollout or integration timetable as of mid‑March 2026. (ign.com) The initiative follows prior Sony actions to curb low‑effort AI content on its platforms, including a removal wave of shovelware and “AI slop” from the PlayStation Store that was reported in early 2025. (eurogamer.net)
