Overview
- Public Citizen on Nov. 11 called on OpenAI to pull Sora 2, citing product safety risks, misuse of people’s likenesses, and threats to democratic trust, and also notified members of Congress.
- AI security firm Mindgard reported on Nov. 12 that Sora 2’s hidden system prompt can be reconstructed via audio transcripts, revealing internal safety guidance.
- Testing by 404 Media showed Sora 2’s filters can be bypassed to generate copyrighted characters and close likenesses of real people despite recent policy changes.
- Hyperrealistic clips tied to current political narratives have circulated with little or no labeling, with outlets such as Fox News initially treating some as authentic before updating coverage.
- Sora remains widely available and popular on iOS and Android, as OpenAI rolls out post‑launch restrictions and secures select agreements with rights‑holders including Martin Luther King Jr.’s family and Bryan Cranston/SAG‑AFTRA.