Meta’s standalone AI application has ignited a firestorm of criticism, creating what many are calling a privacy disaster by publicly broadcasting users’ ostensibly private and often deeply personal conversations for anyone to see. The app’s “Discover” feed, intended as a social feature, is instead exposing a continuous stream of sensitive user queries on everything from legal troubles to medical ailments, in a fiasco that one writer for TechCrunch described as “the start of a 21st-century horror film: Your browser history has been public all along, and you had no idea.”
Launched in April, the Meta AI app allows users to optionally share their chatbot interactions. However, the torrent of publicly visible posts suggests many are completely unaware they are publishing their private thoughts. As reported by Wired, these shared chats include a 66-year-old man asking, “what counties [sic] do younger women like older white men,” while publicly revealing his age and willingness to relocate.
While Meta claims that chats are private by default and only become public after users complete a multi-step sharing process, the public feed is rife with evidence of user confusion. Privacy experts have found the situation “incredibly concerning,” explaining that it reveals a fundamental public misunderstanding of how modern AI systems and data privacy work. This design choice has created a privacy nightmare where users’ names and profile photos are attached to their most sensitive inquiries.
Flawed by Design: A Social Feed of Private Thoughts
The core of the issue lies in the app’s social-by-design “Discover” feed. When the app launched, Meta’s VP of product, Connor Hayes, said the feed was intended to “demystify AI” by showing what it could do. In practice, it has become a public ledger of private moments, drawing comparisons to the infamous 2006 AOL data leak detailed by The New York Times, where pseudonymized user search histories were published.
The examples of unintentional oversharing are extensive. Security expert Rachel Tobac found instances of users sharing home addresses and sensitive court details, while other visible queries include users asking for help terminating a renter’s tenancy.
If a user’s expectations about how a tool functions don’t match reality, you’ve got yourself a huge user experience and security problem.
Humans have built a schema around AI chat bots and do not expect their AI chat bot prompts to show up in a social media style Discover feed —… https://t.co/CJ41THUXCE— Rachel Tobac (@RachelTobac) June 12, 2025
A key reason for the confusion may be the lack of any privacy-related prompts or pop-ups when first using the app, placing users directly into the interface without warning that their chats could become public.
The backlash from privacy advocates has been swift. The Mozilla Foundation has formally demanded that Meta shut down the ‘Discover’ feed entirely until proper protections can be implemented, calling for all interactions to be private by default. The problem has become so apparent that some users have begun trolling the platform, asking how to make water bottle bongs or sharing their résumés to ask for cybersecurity jobs.
The Data Engine: Memory, Training, and Future Ads
Beyond the public-facing issues of the Discover feed, the Meta AI app operates on a data-hungry model that has sparked separate but related privacy concerns. The app’s “Memory” feature is on by default, automatically parsing and storing facts from user conversations to personalize future responses. Meta’s own terms of service bluntly warn users: “do not share information that you don’t want the AIs to use and retain.”
According to Meta’s Help Center, this retained data may be used to improve its AI models, and unlike some competitors, Meta offers no direct way for users to opt out of this data harvesting for training. The setup led one expert from the Consumer Federation of America to tell The Washington Post that the disclosures and privacy settings “are laughably bad.”
This practice is now facing significant legal challenges in Europe. Max Schrems, chair of noyb, highlighted the legal precedent, noting that the European Court of Justice has already rejected Meta’s ‘legitimate interest’ claim for advertising. He then posed the critical question: “How should it have a ‘legitimate interest’ to suck up all data for AI training?”
Furthermore, these intimate conversations could soon fuel Meta’s primary business, as CEO Mark Zuckerberg has openly mentioned seeing a large opportunity to show ads within AI interactions.
A High-Stakes Gambit Amid Internal Crises
Meta’s aggressive push of its AI products, despite their evident flaws, is taking place against a backdrop of immense internal pressure and strategic urgency. The company is making colossal investments to catch up in the AI arms race, highlighted by a recent $14 billion deal for a 49% stake in the data-labeling firm Scale AI. The deal installs Scale AI’s founder, Alexandr Wang, at the head of a new “superintelligence” lab within Meta.
This move is viewed by analysts as a massive ‘acqui-hire’ designed to bypass internal development bottlenecks after a severe talent drain saw most of the original Llama research team depart for competitors. The investment follows significant delays in the development of Meta’s next-generation “Behemoth” AI model. The deal’s structure, a 49% stake rather than a full acquisition, is also seen by some as a strategic maneuver to avoid further antitrust scrutiny from regulators.
The partnership also brings Meta into closer alignment with Scale AI’s controversial and extensive work as a key contractor for the Pentagon. Scale AI’s founder has been a vocal proponent of this work, arguing in a recent op-ed that there is “no room for neutrality in the global technology race.”
This fusion of a consumer tech giant with a firm deeply embedded in military AI adds another layer of ethical complexity to Meta’s AI ambitions, which are already shadowed by ongoing lawsuits alleging its models were trained on pirated books.
Ultimately, the Meta AI app’s privacy crisis is not an isolated incident but a direct consequence of the company’s broader strategy. In its high-stakes race to build a dominant, all-encompassing AI ecosystem, Meta is moving at a breakneck pace, fueled by massive investments and a voracious appetite for user data. The flawed “Discover” feed demonstrates that in this sprint for AI supremacy, user privacy has become an acceptable casualty, forcing a difficult trade-off for consumers and inviting inevitable clashes with regulators worldwide.
