Apple’s AI fake news warnings highlight the technology’s misinformation problem
Jaap Arriens | Nurphoto | Getty Images
An artificial intelligence feature on the iPhone generates fake news alerts, fueling concerns about the technology’s ability to spread misinformation.
Last week, a feature recently launched by Apple that summarizes user notifications using AI threw out inaccurately summarized notifications from the BBC News app about the broadcaster’s story about the PDC World Darts Championship semi-finals, falsely claiming that British darts player Luke Littler had won the championship.
The incident happened a day before the finals of the tournament, which Littler still won.
Then, just hours after that incident occurred, a separate notice generated by Apple Intelligence, the tech giant’s AI system, falsely claimed that tennis legend Rafael Nadal had come out as gay.
The BBC spent about a month trying to get Apple to fix the problem. British state television complained to Apple in December after its AI feature generated a false headline suggesting that Luigi Mangione, man arrested after murder of UnitedHealthcare CEO Brian Thompson in New York, shot himself – which never happened.
Apple was not immediately available for comment when contacted by CNBC. Apple told the BBC on Monday that it was working on an update to fix the problem by adding a clarification that shows when Apple Intelligence is responsible for the text displayed in notifications. Currently, generated news alerts appear as if they came directly from the source.
“Apple Intelligence features are in beta and we are continuously working on improvements with the help of user feedback,” the company said in a statement shared with the BBC. Apple added that it encourages users to report concerns if they see an “unexpected notification summary.”
The BBC is not the only news organization affected by Apple Intelligence’s inaccurate summarization of news alerts. In November, the feature sent an AI summary notification falsely claiming that Israeli Prime Minister Benjamin Netanyahu had been arrested.
The error was flagged on social media app Bluesky by Ken Schwencke, a senior editor at the investigative journalism site ProPublica.
CNBC has reached out to the BBC and the New York Times for comment on Apple’s proposed solution to its AI feature’s disinformation problem.
The problem of artificial intelligence misinformation
Apple advertises its own Notification summaries generated by artificial intelligence as an efficient way to group and rewrite news app notification previews into a single alert on the user’s lock screen.
It’s an Apple feature he says it’s designed to help users scan their notifications for key details and reduce the overwhelming barrage of updates that many smartphone users are familiar with.
However, this has resulted in what AI experts call “hallucinations” — AI-generated responses that contain false or misleading information.
“I suspect that Apple will not be alone in its challenges with AI-generated content. We’ve already seen numerous examples of AI services confidently speaking untruths, so-called ‘hallucinations,'” Ben Wood, chief analyst for technology-focused market research firm CCS Insights, said. for CNBC.
In Apple’s case, because the AI tries to consolidate notifications and condense them to show only a basic summary of information, it has strung words together in a way that inaccurately characterizes events — but confidently presents them as facts.
“Apple had the added complexity of trying to compress content into very short summaries, which ended up delivering the wrong messages,” added Wood. “Apple will no doubt be looking to resolve this as quickly as possible and I’m sure rivals will be watching closely to see how they respond.”
Generative AI works by trying to figure out the best possible answer to a question or query inserted by the user, relying on vast amounts of data to train its underlying large language models.
Sometimes the AI may not know the answer. But because it is programmed to always present an answer to user queries, this can result in cases where the AI is actually lying.
It’s unclear when Apple’s solution to the notification collapse bug will be fixed. The iPhone maker said it expects one to arrive in the “coming weeks.”