Meta’s recently-launched AI tool is raising serious concerns about privacy and user awareness after it emerged that some of the most personal conversations with the chatbot are being posted online – not by hackers or leaks, but by users themselves, unknowingly.

The company’s AI offering, integrated into Facebook, Instagram, and WhatsApp, as well as a standalone Meta AI app, includes a “Discover” feed where user interactions with the chatbot are publicly visible – the idea being that it can show others how to interact with the chatbot.

The problem  is that many users appear not to realise that by hitting “share,” they’re publishing chats – including sensitive, embarrassing or even incriminating content – into a searchable, public space.

‘Discover’ Feed Confusion

Meta maintains that user chats are private by default and only appear on the Discover feed if a person actively chooses to share them. The interface includes a pop-up warning before any content is published: “Prompts you post are public and visible to everyone… Avoid sharing personal or sensitive information.”

Yet the design and labelling of the “Share” feature appears to be at the heart of the confusion, with critics claiming the process feels deceptively informal – more like sharing a post with friends than broadcasting it to a global, public audience.

Even voice messages, photos and screenshots have appeared on the feed — not via a data breach, but because users tapped “share” without grasping the consequences.

Adding to the problem is the perception that Meta AI, being embedded in trusted apps like WhatsApp, Facebook and Instagram, offers a private digital assistant experience. But while WhatsApp messages are end-to-end encrypted, queries entered into Meta AI through WhatsApp are not afforded the same protections.

Meta’s Response

Meta has repeatedly emphasised that agency is on the side of the suers.

In an April press release announcing Meta AI’s rollout, the company said, “You’re in control: nothing is shared to your feed unless you choose to post it.”

In reality, many users appear unaware that they are “choosing” to make anything public at all.

The company says posts can be deleted after publication and that full control over visibility is available in account settings. But for users who never intended to publish their conversations in the first place, the damage – and data trail – may already be done.

How to Lock Down Your Meta AI Privacy

If you’re concerned about what you may have shared, or want to ensure your AI chats stay private, here’s how to secure your settings:

On the Meta AI app:

1. Tap your profile icon.

2. Go to “App settings.”

3. Select “Data & Privacy.”

4. Tap “Manage your information.”

5. Enable “Make all your prompts visible to only you.”

On Facebook, Instagram, or WhatsApp:

1. Navigate to Settings & Privacy.

2. Tap “Privacy Center.”

3. Select “AI at Meta.”

4. Click “Submit an objection request.”

5. Choose your platform (e.g., WhatsApp) and complete the form.

6. Delete any prior shared conversations if needed.

If you’ve already posted a chat publicly, you can tap the three-dot menu on the post to delete it individually – or wipe them all by selecting “Delete all prompts.”

Recommended reading

• ‘Top-level’ Cybercriminal Jailed After Porn Site Blackmail Campaign
• Instagram Testing New Tools to Combat Sextortion
• Children First See Violent Online Content at Primary School Age

This isn’t just a glitch. It’s a case study in what happens when product design fails to align with user expectations — and when those failures affect something as personal as your digital footprint.

So, before you treat Meta AI like a private diary or confidante, make sure you’re not also hitting “share.” Because what happens in the chat… might not stay in the chat.