What to know
- WhatsApp’s AI labeling for Channels goes further than admin tags: a string in the code says WhatsApp “may also automatically detect and label AI content” — a detection pass on WhatsApp’s side that nobody has reported.
- The admin side works via touch and hold on a shared update → Add AI content label; once added, the label can’t be removed, and the code frames labeling as required in some countries.
- Everything here comes from in-development code inside WhatsApp beta for Android v2.26.31.1 — none of it is visible in the app yet.
Channels are WhatsApp’s one-to-many broadcast layer: admins post updates that can reach millions of followers, which also makes them a natural place for AI-generated images and clips to circulate with no marker of where they came from. An AI content label that channel admins can attach to their own updates has started to surface publicly — but the label system in the code doesn’t stop at what admins choose to disclose.
Our teardown of the latest WhatsApp beta for Android (v2.26.31.1), diffed against the build before it, shows the more consequential half: WhatsApp reserving the ability to detect AI content and apply the label itself, plus wording that makes labeling an obligation rather than an option in some countries.
WhatsApp may label AI content automatically
The load-bearing line sits in the explainer text: channels must label AI content they share, and WhatsApp “may also automatically detect and label AI content.” That reads as a detection pass running on WhatsApp’s side, layered on top of whatever admins declare — though the strings don’t spell out how the detection would work. The whole system is gated behind a new ai_provenance_label_enabled flag, and it arrives with explainer entries covering what counts as AI content and how labels get applied. WhatsApp ships obfuscated resource names, so the strings below are identified by their resource IDs rather than developer-given keys:
The admin flow: add a label you can’t take back
The manual side is straightforward. After sharing an update, a channel admin touches and holds it to bring up a new Add AI content label action, with confirmations for success and failure. One string makes the commitment explicit: once a label has been added, it can’t be taken off. The new strings point to the flow:
Labeling is mandatory where local rules require it
Three variants of the requirement text cover different situations — “channels in your country”, “channels in some countries”, and one aimed squarely at “channel admins in your country” — each paired with a Learn more link. The wording is written as an obligation, not an option, though the code doesn’t say which countries are covered:
None of this is user-facing yet — the automatic detection, the labels, and the mandatory-country notices all live in unreleased code for now. We’ll keep watching future WhatsApp betas for the moment AI content labels start appearing on channel updates, and for any sign of the auto-detection switching on.
Leave a Reply