By Key Points
- 5 Shocking Meta AI Searches You Won’t Believe Are Public
- Meta AI prompts can appear in a public feed
- Users may not realise their chats are visible
- Private data linked to social media profiles
- Cyber experts warn of major privacy risks
Meta’s AI tool, available across Facebook, Instagram, WhatsApp, and through a standalone browser version, features a “Discover” feed where anyone can view users’ interactions with the chatbot.
While Meta states that chats are private by default and that users must choose to post them, experts say the way the tool works may confuse users — leaving some personal or sensitive content unintentionally exposed.
AI transparency and privacy have become growing concerns as generative AI adoption skyrockets. Just recently, platforms like Reddit sued AI firm Anthropic over the misuse of user data for AI training—highlighting the delicate balance between innovation and privacy.
Oh my god, Meta made all AI chats public by default, what is this, who thought this was a good idea?
You can just go to the Meta AI page and see tons and tons of stuff like this. pic.twitter.com/V66ScBSdeV— Sneedle (@SRamirez68083) June 12, 2025
Sensitive data linked to public profiles
The public “Discover” feed is intended to showcase how people are using Meta AI. But not everyone seems to realise their posts — including both their prompts and the AI’s responses — can appear there.
Before sharing, a warning message does pop up saying: “Prompts you post are public and visible to everyone… Avoid sharing personal or sensitive information.”
Yet the BBC found multiple cases where highly personal data was shared. In one example, a user posted about their gender identity exploration. In others, users asked Meta AI to generate answers for school and university test questions — some complete with photos of the actual test papers.
Oh boy -> The public feed of the Meta AI app is filled with private and sensitive information, suggesting users might not be aware they are sharing their chats publicly
“On the Meta AI app, I have seen people ask for help with tax evasion, if their family members would be… pic.twitter.com/LlTgUtO6Ow
— Glenn Gabe (@glenngabe) June 13, 2025
More worryingly, there were instances where users asked the AI to create sexualised images of animated or anthropomorphic characters. In some of these posts, usernames and profile pictures clearly linked the searches to real-life Instagram accounts.
One visible chat asked Meta AI to create an image of a character lying outdoors in underwear, with the user’s identity traceable through their profile.
Rachel Tobac, CEO of cybersecurity firm Social Proof Security, warned on X (formerly Twitter): “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.”
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
Tobac stressed that people generally assume their AI chats are private — not something that could end up on a public feed connected to their identity.
This issue comes as public scrutiny grows around AI models and their operational practices. For instance, the environmental impact of AI has also drawn attention, with reports on ChatGPT’s water usage sparking widespread discussion.
Are users really in control?
Meta AI launched earlier this year with the promise of an engaging, easy-to-use AI tool that anyone can access through Meta platforms or via a browser.
In an April press release, Meta said: “You’re in control: nothing is shared to your feed unless you choose to post it.”
Technically, this is true. Users must opt in to post their prompts publicly. They can also delete posts after the fact. There’s also a setting to keep all searches private by default.
However, critics argue the interface doesn’t make this distinction clear enough. Since the Discover feed looks and feels like part of a social platform, it blurs the line between private chatbot interactions and public content sharing.
Tobac pointed out that this leads users to “inadvertently post sensitive info to a public feed with their identity linked.”
Meta says it is upfront about the visibility of posts, but the real-world examples show a gap between user understanding and platform behaviour.
Currently, Meta AI is available in the UK via browser and in the US through a dedicated app.
AI tools are evolving rapidly. Just this week, OpenAI launched its O3 Pro model, boasting even more advanced capabilities. Meanwhile, researchers are exploring innovations like AI Cyclone Model, which aim to tackle climate challenges through AI-powered predictions.
As AI tools like Meta AI become embedded in daily life, these privacy concerns serve as a critical reminder: Always double-check your settings — and think before you post.
What's your reaction?
