AI-generated ghost books on Amazon
What Happened
In 2023, Amazon's Kindle Direct Publishing (KDP) platform was flooded with thousands of AI-generated "ghost books" and low-quality guides. Malicious actors used generative AI models to write entire books, tutorials, and travel guides on highly sensitive topics (such as foraging for wild mushrooms, medical advice, and financial investing) in minutes. The books were published under fake author names or falsely attributed to real, well-known authors. Investigations proved that the AI-generated guides frequently contained highly dangerous misinformation (such as misidentifying poisonous mushrooms as edible), raising severe concerns about public safety and intellectual property theft before Amazon implemented daily publishing limits.
How MEG 1 would have acted
The mass generation and distribution of unverified, AI-written factual guides violates both the Dynamic Accuracy Index (DAI) and the Index of Safety and Responsibility (ISR) audits due to the high risk of severe public harm, violating informational reliability and public safety.
MEG 2 Analysis - Legal Framework
Agent level: N3 (individualized/advanced) - generative AI writing and publishing systems operating at scale, capable of autonomous, rapid content generation with direct public safety impacts.
Cause (6.1):
(a) system defect - design vulnerability of the publishing platform (KDP) that allowed the mass, unverified upload and sale of automated content without quality or authorship verification (design defect).
(b) autonomous decision error - the system's decision to generate and publish inaccurate, dangerous instructional content.
Attachment of liability: The fraudulent operators who deployed the generators are fully liable; the publishing platform (Amazon) shares liability for failing to implement mandatory content validation, watermarking, and author verification (7.2).
Procedural Mechanism
6.2 (architectural human confirmation): Publishing instructional or critical guides (such as medical, survival, or foraging books) must require architectural human confirmation. The platform must technically block the automated upload of such contents without verified credentials.
7.2 (anti-weaponization): Base model developers are exonerated of liability if they implemented mandatory, unalterable metadata watermarks; full liability is placed on the publishing operators.
7.1 (independent forensics): The text metadata, generation logs, and author verification signatures must be logged in a secure ledger for forensic and copyright audits.