Apple to fix misleading AI notification summaries

Imagine picking up your iPhone, seeing a breaking news alert from the BBC, and reading that a tennis legend just came out as gay, or that a high-profile murder suspect just took his own life.

You would probably assume it was real. But neither the BBC nor the people involved had said anything of the sort. It was just Apple’s AI making things up on your lock screen.

Now, after repeated complaints from the British broadcaster, Apple says it will change how notification summaries are displayed.

Hallucinating the headlines

The trouble started shortly after Apple Intelligence notification summaries rolled out in beta to the UK in December.

The BBC complained that Apple’s AI rewrote a notification about Luigi Mangione, the suspect in the UnitedHealthcare CEO shooting, to falsely suggest the BBC had reported he shot himself.

It only got weirder from there. According to the BBC, Apple’s summaries crowned a PDC World Darts Championship winner before the tournament had even begun, and pushed an alert falsely claiming that Rafael Nadal had revealed he was gay.

The BBC pushed back hard, writing that the summaries do not reflect and in some cases completely contradict the original reporting.

It is not just news alerts, either. Users have spotted other jarring glitches in personal texts, like Apple Intelligence summarizing a friend’s casual message that "that hike almost killed me" as an "attempted suicide."

Apple promises a fix

In a statement to the BBC, Apple said a software update in the coming weeks will further clarify when the text on screen is a summary generated by Apple Intelligence.

Basically, Apple plans to make the AI disclaimer more obvious so users do not mistake a hallucinated bullet point for actual reporting.

Apple also noted that the entire feature is optional. If you do not want an algorithm rewriting your alerts, you can disable it entirely or app-by-app by going to Settings, tapping Notifications, and selecting Summarize Notifications.

Why it matters

Apple’s planned fix feels a lot like slapping a warning label on a leaky pipe. Clarifying that text was rewritten by AI might shield Apple from legal blame, but it does nothing to fix the actual failure of the model.

For users, our take is simple: turn off notification summaries for news apps and messaging right now. Saving two seconds of reading is not worth consuming completely fabricated facts, especially when the stakes are real.

For publishers, this is a nightmare scenario. Outlets spend immense resources fact-checking every word, only for a platform-level algorithm to garble the headline on the lock screen and risk destroying the publisher’s credibility.

Apple clearly wanted to prove that Apple Intelligence could be practical right out of the box. But by letting an imperfect model summarize high-stakes, nuanced text, it ended up showing just how fragile on-device generative AI still is.