Meta Removed Muse Image’s Instagram @-Mention Feature, Not the Model

Last updated: July 13, 2026

What Meta’s updated launch page and the linked opt-out setting establish, and what the public record still does not answer.

For how the current image models compare on output rather than consent, see our guide to the best AI image generators.
For performers and creators, this sits inside the broader fight over AI and performers’ likenesses.
The pattern is not unique to Meta. OpenAI faced similar criticism over an opt-out likeness feature in its Sora app before changing course.

A model and a separate identity-reference feature

Meta announced Muse Image on 7 July 2026 as the first image-generation model from Meta Superintelligence Labs. Meta’s newsroom page describes it as available through Meta AI, able to generate images, edit existing ones, combine references, and power creative tools across Meta apps.

The disputed part was narrower. Meta’s launch page said users could @-mention Instagram accounts in the Meta AI app, and that tagging a username let Meta AI use public photos from that profile to build a new image. According to Meta as reported by RNZ, the feature applied to public accounts held by users over 18, and it was turned on by default, so a user had to opt out. Reporting by outlets including Deadline, Variety, and Axios recorded objections from SAG-AFTRA, CAA, and the advocacy group Public Citizen.

On 10 July, Meta added an update to the same page. It said the company had heard feedback that the feature had missed the mark, and that the feature was no longer available. The update referred specifically to generating images by referencing public Instagram accounts. It did not say Muse Image had been withdrawn.

That distinction is the correction this article exists to make. Some headlines and posts described the episode as Meta killing its new image generator. Meta’s page supports a narrower reading: the company removed one Instagram account-reference route and continued to describe the model and its other surfaces.

Meta’s own page preserves both sides of the reversal

The strongest documentary evidence here is not a critic’s statement. It is Meta’s own updated launch page, read on 13 July 2026.

At the top, the page says the public-account reference feature is no longer available. Further down, the original “Rooted in Your World” section still explains how users could @-mention Instagram accounts and use their public photos in generated images. The launch pitch and the withdrawal notice sit in one document.

This does not prove the feature stayed operational. It shows Meta updated the page by adding a note rather than rewriting the launch copy. For anyone reading later, the page state matters: the headline and most of the body still describe a wider product than the one Meta said remained after 10 July.

What the opt-out setting actually was

Meta’s launch page linked to an Instagram Help Center control. Future Stack Reviews confirmed that link points to Instagram’s content-reuse help page. The page’s full text was not retrievable directly, so the wording below is drawn from reporting that quoted the setting and the support document.

RNZ and Macworld reported the same toggle text: “Allow people to reuse your content on Instagram and with AI features.” Macworld quoted the Help Center document, which says other users may reuse published photos, videos, and reels in features like remix, sequence, templates, and stickers, and, in addition, may create content with your Instagram content using AI features at Meta. Macworld also reported that the “Share and Reuse” option has existed on Instagram for several years, and that the support document is broader than Meta’s launch framing, since it describes anyone reusing public content for AI without tagging the account.

Two things follow, and a third does not. The opt-out was not a dedicated, standalone AI control; it was an existing “Share and Reuse” setting that bundles ordinary content remix with AI reuse under one toggle. And the toggle’s own text names AI, so it is not accurate to call it a purely non-AI control. What the reviewed sources do not establish is when the “AI features” wording was added, what the historical default was, and whether the setting reaches Meta’s model training rather than only reuse by other users. Future Stack Reviews confirmed the page link first-party but relied on secondary quotation for the body, and did not archive the historical version of the page.

Account control and person-level permission are not identical

The following is an inference from how the control was structured, not a statement Meta made.

The setting operated at the level of the Instagram account. The reviewed sources do not document a separate control for other people who appear in an account’s public posts. A public account can contain employees, customers, collaborators, family members, or event attendees who do not run it. An account setting expresses the account operator’s preference. It does not, by itself, establish the intentions of everyone visible in the account’s content.

This is a governance point, not a claim about the system’s internals. Meta has not published, in the sources reviewed here, the image-selection or identity-matching process the feature used, so this article does not describe one.

Referencing a photo is not the same as training on it

Some reactions described public accounts as material for training Meta’s models. The public launch materials do not establish that.

The feature used public photos as references when producing an image. Training a model is a separate process that can involve pretraining, fine-tuning, preference optimization, or later reuse of prompts and outputs. Evidence of one generation does not show that a referenced photo entered model weights or a future training set.

The opposite conclusion is also unsupported. Meta’s removal notice does not say referenced photos were excluded from training, deleted, or covered by a broader data objection. The setting’s wording addresses reuse and AI content creation by other users. It does not, on the reviewed sources, map to model training.

So the accurate statement is narrow: the reviewed sources establish reference use. They do not establish training ingestion, and they do not establish that the opt-out controlled training. No product-specific dataset-lineage statement was located in the reviewed sources.

What changed, what did not, what is unknown

What the reversal resolved

Meta’s update resolved one product question: Meta said the public-account @-mention feature was no longer available.

It did not establish the exact time the feature was disabled across Meta’s systems, whether every affected user saw notice before or during use, how prior generations were retained or shared, whether the setting affected any use beyond reference generation, whether the feature was formally withheld from the EU or EEA (it was not reported as available there, but no statement confirms an exclusion), or how the control treated people depicted in an account they did not run. Meta said it removed the most visible feature quickly. The public record still requires readers to keep four things apart: the model, the feature, the account setting, and the unresolved data-use questions.

For a public-account manager, the practical step is verification rather than a fixed instruction. The feature is described as withdrawn, and the current state of the “Share and Reuse” setting has not been reconfirmed here, so any settings walkthrough should be checked against the live app before it is relied on.

FAQ

Methodology and sources

Future Stack Reviews did not install, run, or opt out of anything in Muse Image. This analysis is based on public records read on 13 July 2026.

Primary Meta pages were opened directly: Meta’s Newsroom launch page, which carries both the 7 July launch and the 10 July removal note, and Meta’s AI research post on the model. The Instagram Help Center control that Meta links to could not be extracted directly; its wording is taken from reporting that quoted it. Vendor pages are treated as Meta’s statements, not independent verification. Positions attributed to SAG-AFTRA, CAA, and Public Citizen, and Meta’s statement on eligibility and notification, are drawn from named reporting and from Public Citizen’s own published statement (citizen.org), and are presented as those parties’ claims. Social-media reactions are treated as unverified signal and are not part of the factual core. This article states no legal conclusion. Volatile facts should be rechecked on the day of publication, and Meta’s launch page should be archived to preserve the state described in Section 2.

Sources, accessed 13 July 2026: Meta Newsroom, “Introducing Muse Image” (launch, the @-mention feature, the 10 July update and “missed the mark” wording, the retained feature copy, pricing wording, US and WhatsApp rollout). Meta AI, “Introducing Muse Image and Muse Video” (the model; the @-mention feature and opt-out are not on this page). Instagram Help Center content-reuse page linked by Meta (URL confirmed first-party; body via reporting). RNZ and Macworld (the “Share and Reuse” toggle text, the support-document wording, and that the setting predates Muse Image). Axios and Variety (Public Citizen, CAA, and SAG-AFTRA statements, and eligibility, all attributed), and Public Citizen’s own statement at citizen.org. Eligibility, the opt-out default, non-notification, and the “Share and Reuse” setting are corroborated across additional coverage read for this briefing, including Bitdefender, Techwalla, and Memeburn.

The Decoder, Trending Topics, and the BBC (as reported via ResultSense) described the launch as US-first for Instagram and treated EU and UK availability as not yet live and AI-Act-contingent; Meta announced no GDPR-specific adjustments.