Meta AI App Faces Major Privacy Backlash as Private Chats Leak Publicly

What you need to Know

• Meta launched a standalone Meta AI app in April 2025 powered by Llama models.

• The app centralizes AI chatbot features previously available on Messenger, Instagram, and WhatsApp.

• A serious privacy flaw has led to hundreds of users’ private chats being published publicly.

• The app’s “Share” button defaults to public sharing with no clear warning to users.

• Critics say Meta’s attempt to make AI interaction social was irresponsible and poorly thought out.

---

In April 2025, Meta introduced its standalone Meta AI application, with the goal of consolidating features that were once spread across Messenger, Instagram, and WhatsApp. As a one-stop AI assistant, the application combined writing assistance, image creation, and recall of user preferences based on Meta's internal Llama language models. Yet what began as a futuristic AI sidekick has now become a contemporary digital privacy horror show. Leaks show that hundreds of private user dialogues, such as audio messages, intimate images, and very personal information, have been accidentally made available to the public.

The source of all the controversy is Meta AI's innocuous-looking "Share" button. When users interact with the chatbot, they have the option to share their queries or resultant responses. The interface will show a preview screen prior to posting but here lies the twist: most users don't realize that pressing "Share" posts their conversations into a publicly viewable feed. This has caused millions of text prompts, audio recordings, and pictures to be shared across the web, unwittingly.

Image Credits: Meta AI (screenshot)

Think waking up to recordings of individuals talking about anything from the science of farting to criminal confessions and discovering they did not intend to release them. Perhaps that is the strange new world on the Meta AI feed today. Discussions about mental health matters, medical diagnoses, tax evasion, and even sexuality have been released with their full names, usernames, and sometimes voice messages or home addresses. The material is subject to public comments, with some serious questions regarding harassment, identity theft, and emotional distress.

This colossal privacy violation is the result of subpar UX design. With ChatGPT or Google Gemini, links default to being private and share settings have to be explicitly set. However, Meta AI's sharing defaults to public by default, with merely a tiny, easy-to-overlook disclaimer. This opacity blindsided users. Users found out their conversations were public only after strangers cautioned them. It is a design fail of epic proportions, according to privacy specialists, and they're correct.

If Meta had made clear privacy signals available, or better still, made private sharing the default, this would not have happened. The app doesn't indicate where your shared material is headed or what privacy settings are in effect. If your Instagram is public and connected to Meta AI, then your shared chats are public as well whether they're discussing legal advice, medical symptoms, or humiliating personal searches.

Security researchers and tech journalists have already compiled threads of publicly posted, clearly private conversations on platforms like X (formerly Twitter). These include:

• Home addresses and sensitive court details

• Cheating confessions and legal dilemmas

• Sexual health concerns

• Inquiries about white-collar crimes

• Users asking for help writing legal documents with real names

• Resume sharing alongside queries about cybersecurity jobs

One especially jarring example includes a man with a Southern accent asking, “Hey Meta, why do some farts stink more than others?” This might be amusing at first glance, but when placed next to other serious disclosures, it paints a picture of a digital space where personal information is weaponized for entertainment.

In spite of the outcry, Meta has not made a public apology or taken conspicuous measures to remedy the issue. Users can only safeguard their data by manually altering their privacy controls: going to Settings > Data and Privacy > Manage Your Information, and turning off prompt history and suggestions.

But this hack is not sufficient. For an AI app from a firm that spent billions building AI technology, so meager a set of proactive safety precautions is shocking and unacceptable. It's not merely an oversight of design it's a violation of trust.

Meta's goal could have been to socialize AI interaction and social sharing. But, transforming a chatbot app into a pseudo-social network without proper user training is a move many now consider irresponsible. There's a reason Google never socialized Search into a social feed, and why earlier examples like AOL's 2006 data breach created similar outrage. Making private conversations involving humans and AI go public was never a good idea. The fact that this sensitive posting went viral, combined with Meta's deafening silence, threatens to snowball into a public relations nightmare of gargantuan proportions.

Image Credits: Meta AI (screenshot)

Per Appfigures, the Meta AI app has been downloaded 6.5 million times since its debut on April 29, 2025. While impressive for a new indie app, for Meta a company with billions in R&D it's disappointing. By contrast, the current privacy scandal has received much more publicity than the app's features ever did.

From resumes requesting cyber work, to Pepe the Frog profiles sharing dope-making instructions, the feed is rapidly being taken over by trolls, nosy users, and even journalists chronicling the fiasco. This is not the AI-facilitated productivity Meta had in mind this is a data privacy circus.

Meta's AI ambitions have gotten it into hot water once again. This is not a technical glitch scandal it's one of neglecting user safety and consent in the era of generative AI. While AI technologies are moving to the center of our online existence, this episode shows that privacy-first design is no longer a choice but a must.

If Meta is serious about competing in the AI arena and restoring trust, it needs to rethink its sharing model now, provide explicit user warnings, and default to private engagement. Until it does that, the Meta AI app might be better remembered for its privacy faux pas than for technological advancements.