Meta is testing a Meta AI account on Threads in five countries โ Argentina, Malaysia, Mexico, Saudi Arabia, and Singapore โ that users can tag (like Grok-on-X) to get answers to questions inside conversations. Users quickly discovered they can't block the new account. The block option is either missing or returns errors when invoked. "Users cannot block Meta AI" trended on Threads with over one million posts before the trend appeared to be suppressed in some users' feeds, per Engadget's Karissa Bell. Meta spokesperson Christine Pai told The Verge users can "mute or hide Meta AI replies, or use the 'Not interested' option on any Meta AI post" โ but not block.
Block is the standard user-control primitive on social platforms: it removes a specific account from a user's feed, prevents interactions, and signals consent withdrawal. Mute hides without preventing interactions; "Not interested" downranks but doesn't remove. Removing block specifically for the AI account is a design choice that puts Meta AI's presence past where the platform's user-control infrastructure normally lets users opt out. Meta has been investing heavily in AI to catch OpenAI and Google โ launching the Muse Spark model in April and rolling it across its apps and services. The Threads AI account is the visible front of that strategy. The trending-then-disappearing topic is the separate concerning signal alongside the consent issue: algorithmic suppression of organic complaints about a feature is a different bad pattern from the original carve-out, and harder to prove, but Bell's reporting (the trend hit a million posts then stopped showing in her own feed) is the specific anecdote.
The Human Consent Standard launched by Hollywood A-listers (covered earlier today, #825) and the OpenAI wrongful-death lawsuit (#827) sit on the same broader axis as this Meta design: AI presence in spaces users participate in, without explicit user-controllable consent. The Hollywood standard targets the training-data side (don't train on my likeness without opt-in); the runtime side โ what's actually in your feed without an opt-out โ is exactly what Meta's Threads test is exposing. EU DSA already has user-control requirements where block has been treated as a regulated affordance under "user empowerment" provisions; US has no comparable statute yet. For builders shipping AI features into existing social products: respect the block primitive or become the next regulatory case study. The Grok-on-X precedent matters because that's where Meta got the design language โ xAI also makes Grok un-blockable on X. The cross-platform pattern: AI accounts get carve-outs from user-control infrastructure as soon as they ship, users notice and push back, and platforms walk back partway or absorb the criticism. Meta will likely settle in the partway-walk-back position.
The test is in five non-US, non-EU countries โ the standard Meta launch shape (markets with less media scrutiny, refine the feature, then roll wider). If the block-disabled design survives this test and rolls to the US and EU, expect regulatory pushback โ DSA in EU, state AGs in US, and probable FTC interest under existing user-protection authority. For Threads users in the test countries right now: muting Meta AI is the available control, and the "Not interested" tap on each post is the per-item form. For everyone else watching: the design choice is a signal about what Meta thinks user control over AI presence should look like over the next year. If you've been waiting for a concrete "AI in your feed without consent" example to point to in policy conversations, this is it. The companion question worth tracking is whether the trend-suppression report holds up under more reporting โ algorithmic damping of organic complaints is the harder-to-prove pattern and the more serious one if confirmed.