Why you can check Claude-made images but not Claude-made text

AI content provenance is shipping in two speeds, and the gap between them is regulatory rather than engineering.

Media files move first because the public has a tool. Text moves second because, by Anthropic's own statement to CNET, the separate text Detection API is gated to organizations covered by the EU AI Act, not opened to consumers.

The mechanism is regulatory asymmetry in deployment. The file format that carries an easy, browser-checkable credential ships to everyone, while the format that requires a deeper compliance wrapper ships only to the institutions a law names. CNET's hands-on test confirmed the image check works on a sample file. Anthropic says uploaded files never leave the device and only credentials are read. None of that changes the underlying split.

Creators and journalists get a free way to verify Claude-made images and audio. They do not get a way to verify Claude-made text on the same day. Organizations with EU AI Act exposure get a detection lane now; everyone else is queued behind a private preview and a sign-up form.

The line to watch is simple: when the text detector opens past the EU law's addressees, the public will finally have parity. Until then, the asymmetry is the story.