Alex Preston made a choice that ended his association with the world’s most closely watched newspaper at some point during the editing process, between a first draft and a piece that was submitted. He reviewed a book using an AI tool. The tool discreetly copied text from a Guardian review of the same book and incorporated it into his draft without noticing it because it pulls from online sources in the same manner as these systems. Preston failed to notice it. His editors failed to notice it. Someone read it. Every freelancer in the industry should sit up straight after reading that.
Watching Over Her, written by French author Jean-Baptiste Andrea, was the book in question. In January 2026, Preston’s review appeared in the New York Times. Christobel Kent’s book had been reviewed by the Guardian the previous August. It was difficult to explain away the similarities when a Times reader put them side by side. The descriptions of the characters were nearly identical to one another. The final emotional register of the novels, which is where a critic gets paid by summarizing the final meaning of a book, was almost exactly the same in both works, down to the line “a country of contradictions: battered, divided, misguided and miraculous.” An investigation was started by The Times. Preston acknowledged the incident. In the same week, the paper severed its relationship with him.
| Category | Details |
|---|---|
| Freelancer Dropped | Alex Preston — author, journalist, head of advisory at Man Group investment firm |
| Publication | The New York Times — severed ties with Preston in late March 2026 |
| Book Under Review | Watching Over Her by Jean-Baptiste Andrea — Preston’s NYT review published January 6, 2026 |
| How It Was Caught | A Times reader flagged similarities between Preston’s review and an August 2025 Guardian review by Christobel Kent of the same novel |
| Overlapping Language | Near-identical character descriptions and near-verbatim concluding assessments of the novel — drawn in by AI without Preston’s awareness |
| Preston’s Statement | “I made a serious mistake in using an AI tool on a draft review I had written… I am hugely embarrassed.” |
| NYT’s Verdict | “Reliance on AI and inclusion of unattributed work by another writer is a serious violation of the Times’s integrity and fundamental journalistic standards.” |
| Preston’s Track Record at NYT | Six reviews published between 2021–2026; no prior AI use reported in other pieces |
| Irony | Preston authored a piece for Man Group titled “The AI Bubble: Hidden Risks and Opportunities” earlier in 2026 |
| NYT AI Policy | Internal principles require AI use to be vetted by journalists, reviewed by editors, and disclosed to readers with risk-mitigation steps explained |
Preston’s character is what makes this tale so painful. He’s not a content mill worker meeting daily targets or a young writer scurrying for bylines. In addition to being the head of advisory at Man Group, a significant investment management company, he is a six-time published author and a contributor to the Financial Times, the Economist, and the Guardian. Since 2021, he had contributed to the Times. Additionally, he wrote an article for the Man Group website titled “The AI Bubble: Hidden Risks and Opportunities” earlier this year, weeks before this incident was made public. The irony is so awkward that it seems almost too clever to be true. However, it is true, and it illustrates how these tools can surprise even seasoned, considerate individuals.
Regarding its public stance on AI, The Times has been thoughtful. According to its internal policies, work utilizing generative AI must be approved by journalists, examined by editors, and made public when appropriate. “The first principles of journalism should apply just as forcefully when machines are involved” is one of the principles that is worth quoting in context. Although it was written as a standard, that now reads like a warning. Enforcement is what separates the two, and the Preston case shows that enforcement has mostly been reactive, at least at the freelance level. It wasn’t discovered by any pre-publication procedure; rather, it was discovered by a reader.
This is the current location of the larger conflict in journalism. A growing number of writers at all levels of the industry use AI editing tools covertly and covertly. For minor grammar correction, some people use them. Others use them, as Preston seems to have done, to organize a draft, polish language, and smooth out structure, but they don’t fully comprehend how those systems generate their output. In late March 2026, The Atlantic reported that AI seemed to be appearing in major publications‘ opinion pages without being disclosed. Although the Preston case is the most well-known outcome to date, it is unlikely to be the final one.
Observing all of this gives the impression that the regulations controlling the use of AI in journalism are still being developed in real time, primarily in reaction to incidents rather than in advance of them. The Times responded to Preston in a clear and concise manner, and it has the institutional authority to set a standard. However, it’s possible that dozens of cases go unreported for every well-known freelancer caught in this manner because no interested reader happens to have read both pieces. Editors find that uncomfortable, which is likely why more newsrooms are adopting formal disclosure requirements and pre-publication AI detection, even though those technologies are still not perfect. No one was intended to be plagiarized by the machine. That might be the most disturbing aspect of it all.
