new York
CNN
—
Apple is temporarily removing its new artificial intelligence feature that summarizes news notifications after repeatedly sending users error-filled headlines, sparking backlash from a news organization and advocacy groups of freedom of the press.
The iPhone maker’s rare reversal on its much-marketed Apple Intelligence feature comes after the technology produced misleading or outright false summaries of news headlines that appear nearly identical to regular push notifications.
Apple on Thursday rolled out a beta software update to developers who have disabled AI functionality for news and entertainment titles, which it plans to roll out to all users later as it works to improve AI functionality. The company plans to re-enable this feature in a future update.
As part of the update, the company said Apple Intelligence summaries, which users must opt-in to, will more explicitly emphasize that the information was produced by AI, signaling that it can sometimes produce inaccurate results.
Last month, the BBC complained to Apple about the technology, urging the company to remove the feature after it created false headlines claiming that Luigi Mangione, accused of murdering UnitedHealthcare CEO, had committed suicide . On another occasion, three New York Times articles were also summarized into a single push notification, falsely stating that Israeli Prime Minister Benjamin Netanyahu had been arrested.
A BBC spokesperson told CNN in December that “it is essential that Apple addresses these issues urgently, as the accuracy of our information is essential to maintaining trust.” These summaries from Apple do not reflect – and in some cases completely contradict – the original BBC content.
On Wednesday, the AI-powered feature once again incorrectly summarized a Washington Post notification, falsely stating “Pete Hegseth fired; Trump’s Tariffs Impact Inflation; Pam Bondi and Marco Rubio confirmed. None of this is true.
“This is my periodic rant that Apple Intelligence is so bad that today it got every fact about its AI wrong, a summary of news alerts from the Washington Post,” the tech columnist wrote of the newspaper, Geoffrey Fowler. “It’s completely irresponsible for Apple not to disable summaries in news apps until it gets a little better at this area of AI.”
Press freedom groups have also highlighted the dangers that summaries pose to consumers seeking reliable information. Reporters Without Borders calls them “a danger to the public’s right to reliable information on current events” and the National Union of Journalists, one of the largest journalists’ unions. unions around the world, emphasizing that “the public must not be put in a position where they have to doubt the accuracy of the information they receive.” Both called for the removal of AI-based summaries.
Apple’s AI problems are not the first time a developer has had to deal with information manufacturing technology, with popular models like ChatGPT often producing confident “hallucinations.”
Large language models, the technology behind AI tools, are trained to respond to inputs using “a plausible response” to prompts, Suresh Venkatasubramanian, a professor at Brown University who helped co-write the blueprint for the White House for an AI bill of rights. , previously told CNN.
“So in that sense, any answer that seems plausible, whether it’s accurate, factual, made-up or not, is a reasonable answer, and that’s what it produces,” Venkatasubramanian said. “There is no knowledge of the truth there.”
Two years after the launch of ChatGPT, AI hallucinations remain more prevalent than ever. A July 2024 study by Cornell, the University of Washington and the University of Waterloo found that the best AI models still cannot be fully trusted given their propensity to invent information.