Apple is updating Apple Intelligence after its "smart notifications" feature started spreading misinformation by combining news stories or hallucinating details.
A formal complaint by the BBC in December highlighted news alerts, branded with the BBC logo presenting false information as fact. Among them was one claiming Luigi Mangione, the man accused of killing UnitedHealthcare CEO Brian Thompson, had shot himself. Mangione is currently alive.
The iPhone-maker says it will roll-out the update in the coming weeks and will flag when a notification is actually an AI-generated summary.
Apple told the BBC that its Apple Intelligence features were still in beta and constantly being updated. It is currently only available on select devices running iOS 18.2 or macOS 15.2.
What went wrong?
Almost all Apple Intelligence features happen on-device using a relatively small language model. While AI tools like ChatGPT and Gemini have largely tackled the hallucination problem, this is still an issue for some smaller models.
Apple's smart summaries condense the content of notifications, whether from your email, a website or the News App, to provide a clearer overview. This can result in funny, tragic or worrying results.
For Apple News notifications it looks at the headline and body of the article to generate a short summary. Issues tend to arise when it combines multiple stories into one summary, creating confusing or completely wrong headlines.
In future, Apple says the summaries will be more clearly flagged as AI generated.
