90 incidents on record · 2026 Headlights Incident reports by Ellie Harris · Melbourne
10 new this week Library last updated 13 July 2026
← The incident library
HD-INC-087
Media · United States · 2025 · Hallucination & fabrication

The Chicago Sun-Times ran a summer reading list with ten books that did not exist, from a syndicated section a freelancer had built with AI and not checked

By Ellie Harris · Filed Published 18 May 2025

Alleged: Chicago Sun-Times (Chicago Public Media); King Features (a unit of Hearst); freelancer Marco Buscaglia developed or deployed the AI system implicated in this incident. Details are drawn from public reports; parties are presumed innocent of any wrongdoing not established by an official finding.

The Chicago Sun-Times ran a summer reading list with ten books that did not exist, from a syndicated section a freelancer had built with AI and not checked

What happened

It was reported that on 18 May 2025 the print and e-paper editions of the Chicago Sun-Times carried a special section titled “Heat Index: Your Guide to the Best of Summer”, a 64-page insert produced by King Features, a unit of Hearst, and licensed to the paper. Its “Summer reading list for 2025” recommended fifteen titles with short reviews. Ten of the fifteen did not exist. The authors were real, and well known, but the books were invented: a novel described as Isabel Allende’s first work of climate fiction, and a title attributed to the Pulitzer winner Percival Everett, among others, none of which had been written.

It was then reported that the freelance writer who compiled the section, Marco Buscaglia, had used an AI tool to help research it and had not checked what it produced. A later review found the fabrication was not confined to the book list; other parts of the section contained errors of the same kind, including a quoted expert who could not be verified. The Sun-Times said the content had been licensed from King Features and was not created or approved by its newsroom, apologised, and removed the section from its digital edition; the chief executive of its parent, Chicago Public Media, told staff she was sorry for the mistakes, and the journalists’ union called on management to prevent a repeat. King Features said it had ended its relationship with the writer, whose undisclosed use of AI it said breached its policy. The section had passed from a freelancer to a syndicator to a newspaper and into print, and at no point in that chain had anyone checked whether the books it recommended were real.

What an auditable version would have shown

Every recommendation named a book and an author, and each of those is something you can check against a library catalogue in seconds. Run every title on the list against a real catalogue before it goes to print and ten of them come back as not found. King Features said Buscaglia’s use of AI was undisclosed, which means the people who put the section in the paper did not know it was there to be checked. That is the argument for the check travelling with the work, along with a note of whether it was done, as a section passes from writer to syndicator to newspaper: the last people to touch it can see what was verified and what was not.

Where the gap was

Text naming real books and real authors went to print without anyone confirming the books existed. A CitationVerifier checks each name a piece cites, a title, an author, a study, a quote, against a real database before it can ship, so an invented reference is caught while it is being written, not after it is printed. A ConductRecord travels with the piece through the syndication chain and records what came from AI and what was checked, so a newspaper buying a section can see whether its sources were ever verified instead of assuming they were.

What governance should have looked like

Checking whether a book exists is quick and costs nothing; the trouble here was the hand-off, because syndicated material turns up looking finished and it is easy to assume someone already checked it. Best practice would be for anything that cites real sources to carry, through every set of hands, a record of what was checked and against what, so a publisher can confirm the sources are real before it prints them, instead of finding out in public that ten of them are not.

Failure Pattern: AI-generated text with fabricated references was published without anyone checking those references against a real source.

Governance Principle: content that names real-world sources should be checked against those sources before publication, whoever or whatever produced the draft.

The reference implementation of CitationVerifier and ConductRecord is open source. It lives at github.com/saffronandindia/headlights-oss, Apache 2.0 licensed and free to install. The repository is public now.

Sources

The mailing list

Fresh incident reports every week. One email to match.

We add new incidents to the library regularly, and send a single short email each week with what's new. The library stays free and open; this is just how you keep up with it.

No tracking. Unsubscribe in one click.

The record

An auditable system would have produced a signed, tamper-evident record the moment this happened: what the system did, the version that did it, the basis it acted on, and the action taken, and Chicago Sun-Times (Chicago Public Media); King Features (a unit of Hearst); freelancer Marco Buscaglia could have produced it on demand.

This is the record the system as deployed did not produce in a signed, auditable form.

What this teaches
Capture what happened when it happens
What the system did, the version that did it, the basis it acted on, and the action taken, recorded at the moment, not reconstructed after.
Sign it, so no one has to trust the record-keeper
A tamper-evident entry. Edit it later and the signature breaks. The record does not ask for the benefit of the doubt.
Make it verifiable by anyone
A court, a regulator, a customer's lawyer can check the record themselves, without taking the company, or us, at our word.

Headlights summarises publicly reported AI incidents. All summaries are independently written, attributed to their original sources, and intended for research and educational purposes. Allegations are identified as such until established through official findings.

Last reviewed June 2026. This report is based on the sources listed above and reflects information available at the time of review; later developments may not be captured. Where a person is described as charged with or alleged to have done something, that allegation is unproven unless a conviction or a court or regulatory finding is stated. Headlights publishes journalism and commentary, not legal advice.

Want to write back?

Direct to my inbox.

ellie@useheadlights.com →