AI video-generator restrictions can mean service rules, access limits, or laws governing particular conduct. A provenance check, by contrast, looks for clues about where a video came from. That distinction helps make sense of two current facts: OpenAI discontinued Sora’s web and app experiences on April 26, 2026, and its API on September 24, 2026; Google’s SynthID checks look for supported watermarks, not whether every detail in a video is true.

Sora’s web, app, and API have been discontinued

OpenAI’s Sora discontinuation notice gives separate dates for the service’s web and app experiences and its API. The web and app experiences ended on April 26, 2026. The API was discontinued on September 24, 2026.

What a SynthID check can tell you

A watermark is a marker embedded in media; provenance means information about its origin. Google DeepMind says SynthID embeds an imperceptible watermark in AI-generated video and designs it to withstand changes such as cropping, filters, frame-rate changes, and lossy compression. Google names Gemini, Search, and Chrome as ways to check uploaded media for a SynthID watermark. Its SynthID overview describes detection of Google and supported partner watermarks.

A positive result indicates that the check found a supported watermark. It is a clue about provenance, not a universal test of authenticity. A negative result means the check did not find a supported watermark; it does not establish that the video is genuine.

What the TAKE IT DOWN Act covers

The U.S. TAKE IT DOWN Act addresses specified knowing publication of a nonconsensual intimate digital forgery of an identifiable adult through an interactive computer service, subject to the law’s other conditions. Those include harm or intent to cause harm, and provisions concerning public exposure and matters of public concern. The law does not make every AI video depicting an adult a federal crime simply because it was created. The statutory text also says that consent to creating an intimate depiction does not by itself establish consent to publishing it.

The Act sets a separate duty for covered platforms: after a valid removal request, they must remove a qualifying depiction as soon as possible and no later than 48 hours, and make reasonable efforts to remove known identical copies.