Meta's Muse reaches 10M US downloads amid privacy backlash

Muse’s rapid adoption

Meta’s AI companion Muse hit 10 million U.S. downloads in just 40 days. The milestone puts the app in a narrow club of eleven other applications that have moved that many installs in a comparable window, according to TechCrunch.

The same report notes that Muse is retaining users at higher rates than its AI rivals. Retention metrics were not broken out, but the headline suggests a stickier experience than the chatbots offered by OpenAI, Google, or Anthropic. The numbers come from internal data that Meta shared with the outlet, confirming the download count without inflating the claim.

The surge follows a broader push by Meta to embed AI assistants across its product suite. Muse appears as a standalone iOS and Android app, marketed as a personal companion for brainstorming, scheduling, and casual conversation. Its rapid uptake signals that a sizable segment of U.S. consumers is willing to experiment with conversational agents outside the traditional social network environment.

Privacy slip: sharing feature breach

In early August, Ars Technica reported that Muse exposed users’ personal and business secrets when they used the conversation‑sharing feature. The breach surfaced after a user inadvertently shared a chat that contained confidential client information, prompting a cascade of similar disclosures.

Meta responded by suspending the sharing function while it investigates the root cause. The company has not disclosed how many users were affected, but the incident underscores the risk of treating AI chats as public content. Privacy advocates argue that the default design should have treated shared snippets as private unless the user explicitly opted in.

The episode arrives at a moment when regulators are tightening scrutiny of AI data practices. While Meta has not faced formal enforcement action over Muse, the privacy slip adds pressure to its broader AI compliance agenda.

Policy retreat: TOS changes and internal blame game

Meta’s commercial Terms of Service previously required developers to “train” their models on scraped data. After a wave of criticism from the AI community, the company removed the word “train” from the TOS, as reported by Ars Technica. The edit was framed as a response to “AI backlash,” but the revised language still omits any reference to respecting robots.txt directives.

Simultaneously, internal emails leaked to the press reveal that Meta attempted to distance executives from the decision to use pirated books for training. The company argued that no employee “decided” to download torrents, effectively scapegoating lower‑level staff. The narrative was intended to shield senior leadership from liability, but it also exposed a disconnect between policy statements and operational practices.

These moves illustrate a pattern of reactive policy adjustments rather than proactive governance. By stripping “train” from the contract without adding concrete safeguards, Meta leaves the legal gray area around data scraping largely untouched.

Industry reaction: documentaries, hype, and skepticism

The controversy surrounding Muse coincides with a broader cultural moment. Netflix released a documentary titled AI Gone Rogue that dramatizes the risks of unchecked AI development. Clem Delangue, co‑founder and CEO of Hugging Face, appears in the film and later told TechCrunch that the documentary went “full doomer mode” and is not accurate.

Delangue’s comment reflects a growing frustration among AI researchers who feel that popular media amplifies fear without grounding it in technical nuance. The same sentiment appears in the coverage of Muse: the app’s headline‑grabbing download numbers mask unresolved privacy and data‑use questions. Critics argue that the hype around rapid adoption can obscure the need for robust safety mechanisms.

Both the documentary and Meta’s product rollout demonstrate how quickly narratives can diverge from the underlying facts. While the public sees a flashy AI companion, the engineering teams grapple with data provenance, user consent, and the legal ramifications of large‑scale scraping.

What to watch

The next weeks will reveal whether Meta reinstates the sharing feature and how it amends its Terms of Service to address robots.txt compliance. Analysts will also track retention trends beyond the initial launch window to see if the early adoption translates into sustained engagement. Finally, the industry will watch for any regulatory response to the privacy breach, which could set a precedent for how conversational AI apps handle user‑generated data.