Meta AI's new chatbot raises privacy alarms
If you've ever typed something sensitive into Meta AI, now is the time to check your settings and find out just how much of your data could be exposed.
Sign up for my FREE CyberGuy ReportGet my best tech tips, urgent security alerts, and exclusive deals delivered straight to your inbox. Plus, you'll get instant access to my Ultimate Scam Survival Guide - free when you join my CYBERGUY.COM/NEWSLETTER.
Meta's AI app, launched in April 2025, is designed to be both a chatbot and a social platform. Users can chat casually or deep dive into personal topics, from relationship questions to financial concerns or health issues.
What sets Meta AI apart from other chatbots is the "Discover" tab, a public feed that displays shared conversations. It was meant to encourage community and creativity, letting users showcase interesting prompts and responses. Unfortunately, many didn't realize their conversations could be made public with just one tap, and the interface often fails to make the public/private distinction clear.
The feature positions Meta AI as a kind of AI-powered social network, blending search, conversation, and status updates. But what sounds innovative on paper has opened the door to major privacy slip-ups.
Privacy experts are sounding the alarm over Meta's Discover tab, calling it a serious breach of user trust. The feed surfaces chats containing legal dilemmas, therapy discussions, and deeply personal confessions, often linked to real accounts. In some cases, names and profile photos are visible. Although Meta says only shared chats appear, the interface makes it easy to hit "share" without realizing it means public exposure. Many assume the button saves the conversation privately. Worse, logging in with a public Instagram account can make shared AI activity publicly accessible by default, increasing the risk of identification.
Some posts reveal sensitive health or legal issues, financial troubles, or relationship conflicts. Others include contact details or even audio clips. A few contain pleas like "keep this private," written by users who didn't realize their messages would be broadcast. These aren't isolated incidents, and as more people use AI for personal support, the stakes will only get higher.
If you're using Meta AI, it's important to check your privacy settings and manage your prompt history to avoid accidentally sharing something sensitive. To prevent accidentally sharing sensitive prompts and ensure your future prompts stay private:
On a phone: (iPhone or Android)
On the website (desktop):
Fortunately, you can change the visibility of prompts you've already posted, delete them entirely, and update your settings to keep future prompts private.
On a phone: (iPhone or Android)
On the website (desktop):
If other users replied to your prompt before you made it private, those replies will remain attached but won't be visible unless you reshare the prompt. Once reshared, the replies will also become visible again.
On both the app and the website:
This issue isn't unique to Meta. Most AI chat tools, including ChatGPT, Claude, and Google Gemini, store your conversations by default and may use them to improve performance, train future models, or develop new features. What many users don't realize is that their inputs can be reviewed by human moderators, flagged for analysis, or saved in training logs.
Even if a platform says your chats are "private," that usually just means they aren't visible to the public. It doesn't mean your data is encrypted, anonymous, or protected from internal access. In many cases, companies retain the right to use your conversations for product development unless you specifically opt out, and finding that opt-out isn't always straightforward.
If you're signed in with a personal account that includes your real name, email address, or social media links, your activity may be easier to connect to your identity than you think. Combine that with questions about health, finances, or relationships, and you've essentially created a detailed digital profile without meaning to.
Some platforms now offer temporary chat modes or incognito settings, but these features are usually off by default. Unless you manually enable them, your data is likely being stored and possibly reviewed.
The takeaway: AI chat platforms are not private by default. You need to actively manage your settings, be mindful of what you share, and stay informed about how your data is being handled behind the scenes.
AI tools can be incredibly helpful, but without the right precautions, they can also open you up to privacy risks. Whether you're using Meta AI, ChatGPT, or any other chatbot, here are some smart, proactive ways to protect yourself:
1) Use aliases and avoid personal identifiers: Don't use your full name, birthday, address, or any details that could identify you. Even first names combined with other context can be risky.
2) Never share sensitive information: Avoid discussing medical diagnoses, legal matters, bank account info, or anything you wouldn't want on the front page of a search engine.
3) Clear your chat history regularly: If you've already shared sensitive info, go back and delete it. Many AI apps let you clear chat history through Settings or your account dashboard.
4) Adjust privacy settings often: App updates can sometimes reset your preferences or introduce new default options. Even small changes to the interface can affect what's shared and how. It's a good idea to check your settings every few weeks to make sure your data is still protected.
5) Use an identity theft protection service: Scammers actively look for exposed data, especially after a privacy slip. Identity Theft companies can monitor personal information like your Social Security Number (SSN), phone number, and email address and alert you if it is being sold on the dark web or being used to open an account. They can also assist you in freezing your bank and credit card accounts to prevent further unauthorized use by criminals. Visit Cyberguy.com/IdentityTheft for tips and recommendations.
6) Use a VPN for extra privacy: A reliable VPN hides your IP address and location, making it harder for apps, websites, or bad actors to track your online activity. It also adds protection on public Wi-Fi, shielding your device from hackers who might try to snoop on your connection. For best VPN software, see my expert review of the best VPNs for browsing the web privately on your Windows, Mac, Android & iOS devices at Cyberguy.com/VPN.
7) Don't link AI apps to your real social accounts: If possible, create a separate email address or dummy account for experimenting with AI tools. Keep your main profiles disconnected. To create a quick email alias you can use to keep your main accounts protected visit Cyberguy.com/Mail.
Meta's decision to turn chatbot prompts into social content has blurred the line between private and public in a way that catches many users off guard. Even if you think your chats are safe, a missed setting or default option can expose more than you intended. Before typing anything sensitive into Meta AI or any chatbot, pause. Check your privacy settings, review your chat history, and think carefully about what you're sharing. A few quick steps now can save you from bigger privacy headaches later.
With so much sensitive data potentially at risk, do you think Meta is doing enough to protect your privacy, or is it time for stricter guardrails on AI platforms? Let us know by writing to us at Cyberguy.com/Contact.
Sign up for my FREE CyberGuy ReportGet my best tech tips, urgent security alerts, and exclusive deals delivered straight to your inbox. Plus, you'll get instant access to my Ultimate Scam Survival Guide - free when you join my CYBERGUY.COM/NEWSLETTER.
Copyright 2025 CyberGuy.com. All rights reserved.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
The Verge
an hour ago
- The Verge
Trump says he'll look into deporting Musk as fight over bill escalates
President Donald Trump and Elon Musk's fighting over the 'big, beautiful' domestic policy bill has returned to the spotlight, with the president telling reporters on Tuesday that 'we'll have to take a look' into deporting the billionaire. He also proposed targeting Musk via the Department of Government Efficiency (DOGE), saying, 'We might have to put DOGE on Elon. You know what DOGE is? DOGE is the monster that might have to go back and eat Elon.' Musk has been a longtime critic of Trump's budget bill, which he says he opposes because it will increase the budget deficit. However, a proposed removal of EV tax credits that help Tesla, where Musk is CEO, likely plays a role. After the pair traded insults in early June, both had retreated from publicly squabbling, and Musk deleted some of his posts on X — another of his companies — that linked Trump to Jeffrey Epstein, a convicted sex offender. Then, on Monday evening, as the US Senate worked through a 'vote-a-rama' in an attempt to pass the bill, Musk started posting on X again. He reiterated a threat to primary politicians who support the bill and said, 'If this insane spending bill passes, the America Party will be formed the next day.' Trump posted a response on Truth Social, implying that DOGE (which Musk led before publicly stepping down in May) could cut subsidies for Musk's companies. Without those subsidies, Trump said, 'Elon would probably have to close up shop and head back home to South Africa.' Trump's full response, posted on Truth Social: Elon Musk knew, long before he so strongly Endorsed me for President, that I was strongly against the EV Mandate. It is ridiculous, and was always a major part of my campaign. Electric cars are fine, but not everyone should be forced to own one. Elon may get more subsidy than any human being in history, by far, and without subsidies, Elon would probably have to close up shop and head back home to South Africa. No more Rocket launches, Satellites, or Electric Car Production, and our Country would save a FORTUNE. Perhaps we should have DOGE take a good, hard, look at this? BIG MONEY TO BE SAVED!!! Musk responded on X to Trump's comments, saying, 'So tempting to escalate this. So, so tempting. But I will refrain for now.
Bloomberg
an hour ago
- Bloomberg
Stock Movers: Tesla, Warner Bros., Constellation
Listen for comprehensive cross-platform coverage of the US market close as heard on Bloomberg Television, Bloomberg Radio, and YouTube with Romaine Bostick, Scarlet Fu, Carol Massar and Matt Miller. - Tesla (TSLA) shares fell today as investors expect vehicle deliveries to decline for the second consecutive quarter. CEO Elon Musk has assumed oversight of Tesla's sales in Europe and the US following the departure of longtime deputy Omead Afshar, people familiar with the matter said. Afshar was responsible for Tesla's sales and manufacturing operations in North America and Europe before leaving the EV maker late last month. Following his exit, Musk and Tom Zhu, a senior vice president, are divvying up reporting lines as Tesla looks to recover from another quarter of declining vehicle deliveries. Tesla likely delivered around 389,400 vehicles in the three months that ended in June, according to analysts' estimates compiled by Bloomberg. That would be down roughly 12% from a year earlier, following a 13% drop in the first quarter. - Warner Bros. Discovery (WBD) shares were lower after Advance/Newhouse sold 100 million shares for about $1.1 billion via a block trade, according to a recent filing. - Constellation Brands (STZ) reported earnings and profit trailed expectations in the first quarter due to weaker consumer demand for alcoholic beverages and higher costs from aluminum tariffs .Earnings excluding some items were $3.22 a share in the period, missing the average analyst estimate by 10 cents. The company's beer margins were hit by President Donald Trump's 25% tariff on imported aluminum cans — a key packaging material for its Mexican-made beers such as Modelo and Corona. Constellation is navigating multiple headwinds including tariffs, muted beer sales, and more drinkers cutting back on alcohol and turning to alternatives such as cannabis. The company has also cited a pullback among Hispanic consumers — who make up more than half of Modelo drinkers — amid concerns about inflation, immigration and job security.
Forbes
an hour ago
- Forbes
Granite Demos Show Green Excavation Power For AI Data Centers
Bulldozer on pile of dirtOther construction images: In a canyon accompanied by trailers and vehicles, a few dozen people wearing hardhats are watching a Cat 307.5 excavator at work cutting through hard stone. They're witnessing the Earthgrid company's Plasma Excavation System (PES) where the colossal machine, equipped with two 2500-kilowatt plasma torches, makes a clean border through granite. It's an exciting display, and it represents some of the biggest work being done in renewable tunneling and excavation processes The Environmental Impact Why does this kind of innovation help save the planet and reduce climate impact? For one thing, it makes it easier to put in things like powerlines and utilities underground, decreasing wildfire risks in certain parts of the country that are experiencing severe conflagrations and resulting property damage, as well as harmful emissions. The electric tunneling process replaces dirty diesel equipment, and explosives, which typically have their own impact. There's also the potential to do earth moving in a different way that's less intrusive. All of this combines to make Earthgrid's process very interesting to people who want to improve what we do with infrastructure. CEO Troy Helming was present at Davos talking about the need for such kinds of renewable activity. He's also been vocal about the breakthrough that the company is celebrating now. "(Our recent) developments illustrate a clear step in proving our TBR model and its potential to rapidly connect our underground electrical and fiber optic grid safely. No company has ever tried boring through rock with more than one plasma torch – much less through rock like greywacke or granite," Helming said in a press statement around the process of innovation. "We are excited to deploy our two-torch system on several upcoming customer projects." Coming Up On June 27, the company celebrated a Sierra Granite demo event in Raymond, California. The company chose the spot for its geology and its cache of White Sierra Granite. Helming describes the process this way: 'It's a robot with two lightsabers up front, just sort of vaporizing through rock and soil, with a Mandalorian jet pack on the back, blowing out all the little bits of rock, cornflake-sized bits of rock,' he said. 'And it can go so fast and so much cheaper than anything else out there, it's the power grid, transmission lines, our water pipelines, fiber… there are so many communities that don't have access to highspeed bandwidth. ..' That's one way that the green process will serve our need for energy in a future high-tech age, but here's another: Right now, innovators and big money partners are looking for ways to build out energy sources for large AI data centers. How this gets done is a big question in terms of environmental impact. 'As AI continues to grow rapidly, we're working on even better ways to cool servers efficiently, reduce water usage, and push sustainable, green computing forward,' says Johnson Eung at Supermicro, a top company working on server design. Together, many of these stakeholders will be looking to improve the ways that we usher in the most powerful IT systems ever to exist. That's probably a big deal.
