Hot on the heels of the end of factchecking for Meta, Apple (NASDAQ:AAPL) has temporarily withdrawn its AI-powered news summarisation feature.
The feature was introduced as part of the Apple Intelligence suite but has been suspended following a series of high-profile errors that led to backlash from news organisations, press freedom advocates and, well, truth seekers.
Outright false information
The feature, which generated headlines and summaries for news notifications, had come under fire for producing misleading or outright false information.
In response, Apple deployed a beta software update disabling the feature for news and entertainment notifications.
The company says it plan to refine the technology and reintroduce it in a future update, with added disclaimers emphasising that the summaries are AI-generated and may contain inaccuracies.
The decision follows several high-profile errors. Last month, the BBC raised concerns after AI generated a false headline alleging a suspect in a high-profile case had shot himself.
Another incident saw a push notification falsely summarising three separate New York Times (NYSE:NYT) articles into an incorrect claim that Israeli Prime Minister Benjamin Netanyahu had been arrested.
Similar issues surfaced this week, with the bot inaccurately summarising a Washington Post notification.
Public trust in journalism low
Press freedom groups and journalists – who also have reason to be concerned that AI is replacing them – have warned about the risks posed by AI-generated summaries, particularly in undermining public trust in news at a time when trust in information sources is at an all-time low.
Reporters Without Borders described the feature as a danger to reliable access to current affairs, while the National Union of Journalists underscored the need for public confidence in news accuracy.
Apple’s challenges highlight ongoing issues with large language models, the underlying technology for many AI systems.
Despite advancements, these models remain prone to "hallucinations," or fabricating plausible but false information.
A 2024 study from Cornell, the University of Washington, and the University of Waterloo pointed out this limitation, noting that even top-tier AI systems cannot yet be fully trusted.
Apple's move to pause the feature underscores the challenges tech companies face in deploying AI tools while maintaining accuracy and trust, particularly in sensitive areas like news delivery.