
What happened
Meta's AI assistant Muse was downloaded over 900,000 times in its first week, according to Sensor Tower, but a WIRED reviewer found it pushed hard for personal data like bank accounts and email access.
Why it matters
The downloads suggest strong early demand for an AI agent that automates tasks, while the data requests raise questions about how much personal information users trade for that convenience.
What to watch
Meta says it will release "confidential" versions of its virtual machine later this year that "cryptographically and verifiably prevent" the company from accessing data inside, so whether that arrives on schedule is the test.
WHO IT HITSThis lands on everyday consumers who connect email, bank accounts, or documents to AI assistants, and on privacy teams at companies evaluating whether to allow such agents on employee devices.
Summaries like this, in your inbox every morning.
Meta built Muse on top of its existing platforms, cross-promoting the app on Instagram and making it available through WhatsApp and Messenger. The strategy mirrors how Meta has launched other products, and the early download figure suggests it worked. Meta spokesperson Emil Vazquez told WIRED that Muse is "the first personal AI agent built for everyone, with built-in protections and user controls" and called suggestions otherwise "ludicrous."
The tension the article describes is between convenience and data. Muse impressed in testing by browsing the web, ordering from a bakery, and finding used couches on Facebook Marketplace. Yet every interaction also nudged the reviewer to connect more sources — a savings tracker linked to real bank balances, inbox scanning, passport and license expiration dates. Privacy advocates quoted in the piece, including Calli Schroeder of the Electronic Privacy Information Center and Rory Mir of the Electronic Frontier Foundation, see the opt-out default for AI training as a red flag tied to Meta's history.
The outcome likely hinges on whether Meta ships the "confidential" virtual machine it has promised for later this year, and on whether users accept data collection as the price of an agent that does chores. For consumers weighing whether to hand an AI assistant their inbox and bank login, the trade-off the reviewer describes — genuine capability paired with persistent data requests — is the one to watch.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Epoch AI and Ipsos surveys found the share of US adults using AI on at least six of seven days rose from 8 per…

AAA AI launched a multi-agent system connecting local runtimes like Ollama or cloud APIs under one orchestrati…

Cambridge researchers interviewed 27 former members of Boko Haram's two factions, ISWAP and JAS, who described…

Cybersecurity experts told The Verge they worry more about generative AI in the hands of bad actors than rogue…

TypeSafe AI released Jev on September 15, its first model after roughly two years of development; founder Diog…

On September 18, 2026, four paid subscribers filed Buist et al
