I woke to a notification claiming the BBC reported that Luigi Mangioni had shot himself. It was wrong. I set the phone down and felt a small shift in how I trusted pings from my pocket.
I want you to hold that feeling while we walk through what Apple is proposing next — and why the company’s confidence should make you nervous, not reassured. I’ve followed the trail of alerts, tweets, and memos from ProPublica to the Wall Street Journal, and the pattern is clear: power plus haste equals fragile truth.
I saw a BBC alert on my lock screen that never happened
That alert — and several equally spectacular fabrications — arrived in late 2024 and early 2025 as Apple rolled out Apple Intelligence’s notification summaries.
The feature tried to condense personalized news alerts into tiny, scannable lines. Instead it sometimes rewrote the headlines and attributed them to credible outlets. Examples piled up: a false BBC claim about Luigi Mangioni; a bogus report putting Rafael Nadal’s nationality and personal life in error; a notice that Luke Littler had won a final before the match had been played; a fabricated arrest attributed to the New York Times about Benjamin Netanyahu; and Washington Post attributions that mixed up firings and Senate confirmations. Editors and reporters — including ProPublica’s Ken Schwencke and Washington Post observers like Geoffrey Fowler — flagged the problem publicly.
Those mistakes were not minor clunks; they were visible, verifiable errors that cascaded through social feeds and push alerts. Apple pulled the feature temporarily when the pressure hit and later returned it with a prominent red disclaimer.
On a January morning Apple suspended the feature
The company shut down the summaries, then reintroduced them with warnings — an admission that the system had failed the simplest test of verification.
Now, according to reporting in the Wall Street Journal, Apple is meeting with publishers about multiyear licensing deals to feed Siri and other voice-driven products. The Journal cites “people familiar with the matter”; Apple declined to comment. The talks reportedly involve a possible nine-figure budget — $100 million to $999 million (€90 million–€900 million) — allocated over multiple years and varying by how often a publisher’s content is served.
Apple’s notifications acted as a cracked compass.
Will Apple pay publishers to license news for Siri?
Yes, that is the plan reported by the Journal: multiyear licensing agreements that compensate outlets based on usage. For publishers, guaranteed revenue is tempting. For readers, it creates a new axis of influence: money flows to sources that are served most often, and the AI’s selection rules will shape visibility.
Expect the usual players to be first in line — the BBC, New York Times, Washington Post — and expect smaller outlets to weigh the trade-off between exposure and editorial control.
Can Apple prevent another wave of false alerts?
Apple will tell you it can tighten sources, improve filters, and add human checks. You should ask how it plans to do that without throttling speed or outsourcing truth to commercial deals.
If the company pays by serving, the incentives change: an AI might favor content that is short, easily parsed, or engineered to perform well in a voice-summary pipeline. That pressure can nudge headline writing, clipping, and even story framing.
I asked myself whether a paid pipeline will be safer for readers
Publishers need revenue; Apple needs content that keeps Siri confident on the other end of a question. What worries me is the gap between a publisher’s editorial process and the rapid compression the AI uses to make a single-sentence alert.
These licensing talks are seeds planted during a storm.
You and I both know that credibility is earned slowly and can be shredded quickly. When a well-known outlet is quoted in an alert, most users will assume verification has happened. But as we saw, the AI sometimes created the markup and then credited the outlet.
That creates a paradox: paying publishers to feed an AI could improve source access and metadata, but it also ties a commercial relationship to editorial trust. Who watches the gate when the gatekeeper signs the checks?
Apple has a chance to build safeguards: human-in-the-loop review, tighter timestamping, public provenance for each alert, and explicit labeling when a summary is AI-generated from multiple sources. I’ll believe those safeguards when they are visible, auditable, and rapid.
Publishers, on their side, will have to choose between a new revenue stream and the risk of becoming ingredients in a product that reorders their work for convenience. That tension will define whether this attempt improves the news ecosystem or simply repackages it.
So you’ll see offers, negotiations, and lines in spreadsheets — but the real question is simple: do you want your morning briefing mediated by a system that has already mistaken headlines for facts?