AI models typically learn by memorizing patterns, then researchers bolt on reasoning as an afterthought. A new method called Reinforcement Pre-Training flips this approach—teaching models to think during basic training instead.
Good Morning from San Francisco, Meta's AI app turned users into accidental exhibitionists. 😬 People shared medical questions, legal
Meta users think they're chatting privately with AI. Instead, they're broadcasting medical questions, legal troubles, and relationship problems to the world through a public feed that many don't realize exists.
Meta users think they're chatting privately with AI. Instead, they're broadcasting medical questions, legal troubles, and relationship problems to the world through a public feed that many don't realize exists.
💡 TL;DR - The 30 Seconds Version
🤦 Meta AI users accidentally broadcast private conversations about legal troubles, medical problems, and relationship drama to the world through a public "Discover" feed.
📊 Only 6.5 million people downloaded the app since its April 29 launch, but those who did shared sensitive details including home addresses and court case information.
👴 Older users appear most affected, treating the chatbot like customer service and starting conversations with "Dear Instagram Team" without understanding privacy settings.
🎙️ Voice recordings make the problem worse, with users accidentally sharing audio clips of private conversations, work schedules, and pocket dials.
⚙️ Users can fix this by changing settings to make all prompts private, but Meta's design choices echo past social media privacy disasters.
🚨 The mess shows how rushed AI features can turn users into unwitting exhibitionists, especially when companies don't consider real-world usage patterns.
Meta's AI app has a problem. Users think they're having private conversations with a chatbot. Instead, they're broadcasting their secrets to the world.
The Meta AI app launched in April with a "Discover" feed that shows other users' conversations with the chatbot. People share medical questions, legal troubles, relationship problems, and financial details. Many appear to have no idea their chats are public.
One user asked for help writing a character reference letter for a friend facing criminal charges. They included full names and case details. Another person shared their phone number while asking the AI to post it on Facebook groups to find women to date. A third discussed their sister's potential involvement in corporate tax fraud, naming the specific city and their real Instagram profile.
The accidental confessions range from embarrassing to genuinely concerning. Users have shared:
Voice recordings create an even more uncomfortable experience. Users can accidentally share audio clips of private conversations, shift schedules with coworkers, or pocket dials where Meta AI keeps interrupting real conversations with suggestions.
Meta says conversations aren't public by default. Users must click "Share" and then "Post" to add their chat to the Discover feed. The company insists this is a multi-step process that requires deliberate action.
But evidence suggests many users don't understand what they're doing. One person told a reporter they hadn't intended to make their car repair question public. Others have been warned by strangers in the comments that their sensitive information is visible to everyone.
The confusion may stem from unclear privacy settings. When users log into Meta AI through Instagram, their posts inherit the same privacy settings as their Instagram account. If that account is public, so are their AI conversations.
Venture capitalist Justine Moore noted on X that the feed is "almost entirely boomers who seem to have no idea their conversations with the chatbot are posted publicly."
Some users treat Meta AI like customer service, starting conversations with "Dear Instagram Team" or asking for help logging into their accounts. Others confuse the chatbot with a private assistant, asking it to send reminders about canceling subscriptions or scheduling appointments.
Electronic Privacy Information Center counsel Calli Schroeder calls the situation "incredibly concerning," she told Wired. She points to a fundamental misunderstanding of how AI chatbots work and what happens to user data.
"People really don't understand that nothing you put into an AI is confidential," Schroeder told Wired. "None of us really know how all of this information is being used. The only thing we know for sure is that it is not staying between you and the app."
The problem echoes past social media blunders. Facebook once had a search bar that users confused with the status update field, leading to accidentally public posts. Venmo's default public payment history exposed users' personal details through transaction descriptions.
Meta spokesman Daniel Roberts emphasizes that sharing requires explicit user action. The company's blog post about the app states: "Nothing is shared to your feed unless you choose to post it."
But when asked about mitigations for sharing personally identifiable information, Meta didn't respond. The company also declined to comment on why so many users appear to share content accidentally.
This mess comes as Meta doubles down on AI. CEO Mark Zuckerberg says the company's AI assistant has reached 1 billion users across Meta's platforms. Reports suggest Meta is creating a new AI lab focused on building "superintelligence."
The company has invested billions in AI development. Yet the consumer-facing features often feel underwhelming. Meta tried celebrity-voiced chatbots that flopped. Instagram Messenger hosts user-generated romance bots. Now the standalone AI app creates a feed of accidental oversharing.
As news outlets reported the privacy issues, trolls began posting obviously fake conversations to the platform. Someone shared their resume asking for cybersecurity jobs. Others posted with Pepe the Frog avatars asking about making water bottle bongs.
The trolling makes it harder to distinguish between genuine accidents and deliberate pranks. But the initial wave of deeply personal posts suggests many users genuinely didn't understand the privacy implications.
Users who want to keep their conversations private can adjust their settings. In the Meta AI app, tap the profile icon, then "Data & Privacy" under "App settings." Select "Manage your information" and choose "Make all your prompts visible to only you."
Users can also delete previously shared conversations by selecting "Delete all prompts" in the same menu.
The Meta AI privacy debacle illustrates a broader problem with how tech companies roll out AI features. Companies rush to add AI capabilities without fully thinking through the privacy implications or user experience.
Meta's "Discover" feed concept isn't inherently bad. Seeing how others use AI tools can be educational and entertaining. But the execution creates a minefield of accidental oversharing.
The situation also highlights the digital divide. Younger users who grew up with social media may better understand privacy settings and sharing mechanics. Older users, many of whom joined Facebook to connect with family, may not grasp the implications of AI chatbot interactions.
Why this matters:
Q: How many people have downloaded the Meta AI app?
A: Only 6.5 million people downloaded the Meta AI app since its April 29 launch, according to app intelligence firm Appfigures. That's surprisingly low for one of the world's biggest tech companies launching an AI product it invested billions in.
Q: How exactly do conversations become public on Meta AI?
A: Users must click "Share" then "Post" to add conversations to the Discover feed. Meta says this is a multi-step process requiring deliberate action. However, many users appear confused by the interface and accidentally share private conversations.
Q: Can I make my Meta AI conversations completely private?
A: Yes. In the Meta AI app, tap your profile icon, then "Data & Privacy" under "App settings." Select "Manage your information" and choose "Make all your prompts visible to only you." You can also delete previously shared conversations.
Q: Are voice recordings also being shared publicly?
A: Yes. Users can share audio recordings of their conversations with Meta AI. Some recordings appear to be accidental pocket dials or people who didn't realize they hit the voice button, making the privacy violations even more concerning.
Q: Is this similar to past social media privacy mistakes?
A: Yes. It echoes Facebook's old search bar that users confused with the status field, causing accidental public posts. It also resembles Venmo's default public payment history, which exposed personal details through transaction descriptions until settings were changed.
Q: How does Meta AI compare to ChatGPT or Google's Gemini?
A: Meta's Llama model isn't as advanced as OpenAI's latest models or Google's Gemini Deep Research for complex tasks. It's better suited for entertainment and casual chatting, which explains why the public feed exists at all.
Q: Why are older users more affected by this problem?
A: Many older users treat Meta AI like customer service, starting conversations with "Dear Instagram Team" or asking for login help. They may not understand modern social media privacy mechanics that younger users grew up with.
Q: How big is Meta's overall AI user base?
A: CEO Mark Zuckerberg announced that Meta's AI assistant has reached 1 billion users across all Meta platforms (Facebook, Instagram, WhatsApp). The standalone app's 6.5 million downloads represent a tiny fraction of this total user base.
Disney and Universal sued AI company Midjourney for using their characters without permission to train image generators. It's the first major Hollywood lawsuit against generative AI, testing whether copyright law protects creators in the age of artificial intelligence.
OpenAI cut its o3 model prices 80% while launching o3-pro—a reasoning AI that takes over 3 minutes to respond but outperforms rivals on complex tasks. The move intensifies AI pricing wars and splits the market between fast chat models and slow thinking ones.
Publishers built their business on Google sending them traffic. Now Google's AI answers questions directly, cutting out the middleman. Major news sites lost half their visitors in three years. Some adapt with new revenue models, others fight with lawsuits.
Meta spent months claiming its AI was "crushing it" while losing most of its research team. Now Zuckerberg is spending $10+ billion to fix problems he said didn't exist. The crisis shows how AI forces companies to admit failure or fall behind.
Get tomorrow's intel today. Join the clever kids' club. Free forever.
