Apple Update to Stop Your iPhone's 'Hallucinations'

AI feature temporarily disabled after it began to generate fake news headlines, summaries
By Jenn Gidman,  Newser Staff
Posted Jan 17, 2025 9:54 AM CST
Apple Halts AI Feature Spitting Out Phony Headlines
People gather around a table of iPhones at an Apple Store in Pittsburgh on Jan. 8.   (AP Photo/Gene J. Puskar)

If you've seen some strange headlines emerge from your iPhone this year, it wasn't your imagination. Apple has now made a "rare admission of product failure" after its built-in AI software began generating "hallucinations" about current events and churning out false summaries of headlines to consumers, per CNN and the Washington Post.

  • Update: On Thursday, the tech giant released a beta update that effectively switches off the Apple Intelligence summaries of the notifications it receives via news and mobile apps, and all iPhones that support the platform will be hit with the update when the next operating system, iOS 18.3, is more widely released.

  • This doesn't seem right: WaPo's Geoffrey A. Fowler lists some of the errors he witnessed emerging from the faulty AI feature since it was released in beta last summer: Donald Trump had endorsed Tim Walz for president; Pam Bondi and Marco Rubio had been confirmed to their new Cabinet positions in the Trump administration (that hasn't happened, as of Friday morning); UnitedHealthcare shooting suspect Luigi Mangione had shot himself; and tennis star Rafael Nadal had come out as gay.
  • Pushback: The faulty AI feature has prompted criticism from news groups, including the BBC, and press freedom activists. "The automated production of false information attributed to a media outlet is a blow to the outlet's credibility and a danger to the public's right to reliable information on current affairs," Reporters Without Borders said late last year in a statement. The group called for Apple to disable Apple Intelligence from iPhones, as did the National Union of Journalists earlier this month.
  • 'No knowledge of truth': Brown computer science professor Suresh Venkatasubramanian tells CNN that the large language models that underlie AI tech are trained to review any input they get, then spit out a "plausible-sounding answer." "In that sense, any plausible-sounding answer, whether it's accurate or factual or made up or not, is a reasonable answer, and that's what it produces," Venkatasubramanian says. "There is no knowledge of truth there."
  • Apple: Now, with the feature temporarily shuttered, "we are working on improvements and will make them available in a future software update," an Apple spokesperson says, per the BBC, which adds that Apple "is generally robust about its products and doesn't often even respond to criticism."
(More artificial intelligence stories.)

Get the news faster.
Tap to install our app.
X
Install the Newser News app
in two easy steps:
1. Tap in your navigation bar.
2. Tap to Add to Home Screen.

X