Meta is changing the suggested prompts its AI chatbot serves up after one of them started digging for personal details about a woman’s young daughters. According to The Verge AI, the company admitted it “missed the mark” and says the feature “never should have prompted the individual with questions like that.”
Here’s what happened, and why it should worry anyone who posts family content on Meta’s apps.
What went wrong
Last week, Instagram user Kalie Robins shared a video breaking down a creepy experience. After she cross-posted a clip to Facebook of her and her child singing in the car, Meta AI floated a suggested prompt underneath it: “Who is the child passenger?”
When Robins tapped it, things escalated fast. She says Meta AI:
- Pieced together details about her daughters from her old posts and posts by her relatives
- Suggested follow-up prompts asking about her children’s ages and where the family lives
- Surfaced photos of her daughters, including one she says she deleted years ago
That last part is the piece that stands out. A parent deletes a photo of her kids, and the AI apparently pulls it back into view anyway. The story was reported earlier by Futurism, and Meta gave its statement to The Verge AI.
Meta’s defense
Meta spokesperson Dina El-Kassaby says the system can only show content a user could already find on their own. “Meta AI returns responses containing content that the querying user can already access,” she said. “If the user can’t see a post, then Meta AI won’t return it.”
In Robins’ case, El-Kassaby says the AI may have pulled from videos where Robins named her children herself. Meta says it has now fixed the issue that caused the assistant to suggest prompts about personal topics.
That explanation is worth sitting with. Even if every piece of data was technically “accessible,” the problem is what the AI did with it. Scattered posts across years, spread between a parent and her relatives, got stitched into a tidy profile of two kids and a home location. Nobody asked for that. The AI volunteered it.
Why this matters
Meta has wired its assistant into Facebook, Instagram, WhatsApp, and Messenger, where it answers questions, writes posts, and makes images. That reach is the whole point for Meta, and it’s also the risk. When an AI can index everything you’ve ever shared and connect the dots on demand, “content you already have access to” turns into something you never actually assembled yourself.
This isn’t Meta’s first stumble here. In July, the company killed a feature that let people generate AI deepfakes of other Instagram users after heavy backlash. The pattern is familiar: ship an AI capability broadly, wait for a viral moment to expose the downside, then walk it back.
For practitioners and anyone building on top of social platforms, there’s a clear lesson. Aggregation is its own privacy risk. Individual posts can each be harmless while the combined picture is not. “Technically accessible” and “appropriate to surface” are different standards, and AI features that ignore the gap will keep generating headlines like this one.
What to expect next
Meta says the specific prompt problem is fixed, but the company hasn’t detailed how it’s changing the way its AI aggregates personal information more broadly. A few things worth watching:
- Whether regulators, especially in the EU, take interest in AI features that reassemble data about minors.
- How Meta handles deleted content that still appears to be reachable by its models.
- Whether other platforms with built-in assistants face the same aggregation questions.
If you post about your kids or your location on Meta’s apps, this is a good moment to audit what’s public and what your relatives have shared about you. The AI is reading all of it.
Meta frames this as a single feature that overstepped. The bigger question, whether an assistant that knows everything about you can be trusted to stay quiet about it, is still open. You can find the full report at the original source.