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-037
Technology · United States · 2023 · Hallucination & fabrication

CNET quietly published 77 finance articles written by AI, then corrected more than half of them

By Ellie Harris · Filed November 2022 (reported January 2023)

Alleged: CNET (Red Ventures) 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.

CNET quietly published 77 finance articles written by AI, then corrected more than half of them

What happened

Starting in November 2022, CNET published 77 personal-finance explainers generated by an in-house AI tool. They ran under the byline “CNET Money Staff,” and a reader learned that automation was involved only by hovering over the byline, where a line noted the piece was produced “using automation technology.”

In January 2023 Futurism reported the practice. One explainer on compound interest stated that a $10,000 deposit earning three percent interest a year would earn $10,300 in the first year. That is wrong: the interest earned in the first year is $300, and $10,300 is the total balance, principal plus interest. That prompted an internal review. CNET’s editor-in-chief, Connie Guglielmo, said the review found that 41 of the 77 articles required correction, more than half, with some corrections described as substantial and others minor: incomplete company names, transposed numbers, vague phrasing. Some passages were also found not to be “entirely original,” raising a separate concern that the tool had reproduced text from elsewhere.

CNET paused the AI-generated program and revised how it disclosed automation. It said it intended to keep using AI tools in its work. The failure was not that a tool helped draft an article. It was that factual, money-related claims were published to the public without verification, and that the use of AI was disclosed only to readers who happened to hover.

What an auditable version would have shown

When more than half of a batch of published articles needs correction, the editorial questions are immediate: which claims were checked before publishing, and which parts were machine-generated. CNET could not answer either from the workflow itself; it had to reconstruct the position article by article after the fact. An auditable version records, for each piece, which passages were AI-generated, what factual claims and figures were checked and against what, and who signed off, captured at the time. That record turns a public audit of the whole back catalogue into a query, and it makes honest disclosure a by-product of how the work is logged rather than a separate promise.

Where the gap was

There were two gaps, and they compound. First, factual claims were published without a verification step, in a domain (personal finance) where a transposed number is a real harm to a reader. Second, authorship was not disclosed plainly, so readers could not weight the content accordingly. The controls map to each: a VerificationGate routes factual claims and figures to a trusted source before publication and holds anything that cannot be confirmed; a ConductRecord preserves what was generated, what was checked, and by whom, which is also what makes a clear AI-use disclosure possible. The drafting speed was never the problem. The unverified, undisclosed publishing was.

What governance should have looked like

The lesson is not “do not use AI in a newsroom.” It is that the line between a draft and a published claim is an editorial responsibility, and AI does not move it. Before a factual article ships, its claims are checked against sources, the use of automation is disclosed to the reader plainly rather than on hover, and there is a record of both. That is ordinary editorial diligence, applied to a faster drafting tool, and it is far cheaper than correcting forty-one articles in public.

The reference implementation of VerificationGate 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 CNET (Red Ventures) 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 →