110 incidents on record · 2026 Headlights Incident reports by Ellie Harris · Melbourne
10 new this week Library last updated 27 July 2026
← The incident library
HD-INC-105
Consumer AI · United States · 2023 · Unverified AI-generated content

AI-written mushroom foraging guides sold on Amazon gave dangerous advice under fake authors, and experts warned they could get someone killed

By Ellie Harris · Filed AI-written foraging guides on sale during 2023

Alleged: Amazon (third-party sellers) 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.

AI-written mushroom foraging guides sold on Amazon gave dangerous advice under fake authors, and experts warned they could get someone killed

What happened

It was reported that in 2023 Amazon and other retailers were inundated with mushroom-foraging and identification guidebooks that strongly appeared to be written by AI, sold under author names that could not be traced to any real foragers, and none of them marked as AI-generated. The New York Mycological Society warned on social media that retailers had been flooded with such books, and urged people to buy only titles by known authors and foragers, saying the choice could “literally mean life or death”. Its president, Sigrid Jakob, noted that there are hundreds of poisonous fungi in North America and several that are deadly.

According to reporting, including an investigation by The Guardian, the books carried the formulaic language typical of AI-spun ebooks and contained dangerous advice, such as identifying mushrooms by smell and taste, which is exactly how an inexperienced forager can be poisoned. The Guardian sent four excerpts to Originality.ai, a service that estimates whether text is AI-generated, and all four came back with a 100 percent AI score. Because foraging mistakes can be fatal, mycologists warned the guides posed a real physical risk to beginners who trusted them.

What an auditable version would have shown

A guidebook that tells a beginner which wild mushrooms are safe to eat is safety-critical advice, and the reader has no way to tell a genuine expert from a machine wearing an expert’s name. An auditable version makes two things checkable at the point of sale: whether the content is AI-generated, marked plainly rather than hidden, and whether the safety claims in it were verified against an authoritative source before it was published. With those, a buyer can see that a foraging guide is machine-written and unverified before they carry it into a forest, rather than discovering it the hard way.

Where the gap was

Dangerous, unverified advice was sold under the apparent authority of an expert whose identity could not be verified, with nothing telling the buyer either that it was AI-generated or that its claims had never been checked. A VerificationGate holds safety-relevant claims to evidence before they are published, so “you can identify this mushroom by taste” has to be checked against real mycology rather than generated and sold. A CitationVerifier grounds the content against authoritative sources and flags claims that cannot be supported, so a guide that invents identification advice is caught before it reaches a reader. The failure was not that AI can write, but that AI-written, unverified safety advice was dressed as expertise and sold with nothing standing between it and a beginner in the woods.

What governance should have looked like

When content carries real physical risk, the bar is higher than for an ordinary book, and both the source and the safety of the advice should be visible to the buyer. A machine-written foraging guide should say so, plainly, and its identification advice should be verified before it is ever sold, because the cost of being wrong is not a bad review, it is poisoning. The lesson is that cheap AI content becomes dangerous the moment it puts on the authority of an expert in a field where a mistake can kill.

Failure Pattern: AI-generated, unverified safety advice was sold under fake author identities with no disclosure that it was machine-written, dressing dangerous misinformation as human expertise.

Governance Principle: content that carries real physical risk should disclose when it is AI-generated and have its safety claims verified against authoritative sources before it is sold.

The reference implementation of VerificationGate and CitationVerifier 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 Amazon (third-party sellers) 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 →