India: 70,000 AI-Suspect Books Found in Tamil Nadu Government Purchases Worth About 60 Million Rubles
Writer Perumal Murugan found books in Tamil Nadu's new state library, bought with public money, that contain invented facts and fake illustrations and look like AI-generated text. At first about 1,800 titles from a single publisher were under suspicion, but statewide audits pushed the count of questionable books to 70,000, worth 70 million rupees (about 60 million rubles). Four committees let the purchases through because they checked paperwork, not content. The takeaway: read what you buy, at least in samples.
AI-processed from TG: @hiaimedia (Hi, AI • Новости технологий); edited by Hamidun News
In the Indian state of Tamil Nadu, thousands of books were bought with public money for a new library, and some of them, according to a well-known writer, were written by a neural network. At first about 1,800 titles from a single publisher came under suspicion. Then statewide audits raised the number of questionable books to 70,000. Their cost is about 70 million rupees, roughly 60 million rubles. An important caveat: "questionable" does not mean "proven to be written by AI." Investigators still have to establish what share of the books was generated.
How it came to light
Perumal Murugan is the first Tamil author to be shortlisted for the International Booker Prize. He said that some of the books for the new library were written by AI. Imagine opening a reference book and finding invented dates, people who never existed, and pictures that could not exist in nature. That is what Murugan found: errors, made-up facts that look like neural network hallucinations, fake illustrations, and plainly fictitious author names.
All 1,800 titles in the first batch came from one publisher, Agal Vilakku. Murugan says that in 40 years in literature he had never heard of it. He allows that the company may have been created solely for fraudulent supply deals. Even so, four committees approved the purchase. Formal checks ran like an assembly line: the paperwork was in order, so the books were deemed fit. Nobody opened the books themselves.
What the authorities did
The anti-corruption agency opened a criminal case and froze Agal Vilakku's accounts. The publisher's owner was arrested, and five officials were suspended from duty.
The scandal triggered a review of procurement across the state. It uncovered more than 70,000 questionable books worth about 70 million rupees. The state court called the case exceptional and ordered the authorities to report on all book purchases over the past 20 years. State School Education Minister Rajmohan Arumugam called the affair "the world's first AI book scam."
One of the sources the post relies on describes the purchase as a "30-day fast track." Haste usually weakens oversight, but whether that caused the failure is something the investigation will have to prove.
What this means for practitioners
There is nothing technically sophisticated in this story. It required a cheap stream of plausible-looking text and a procurement system that looks at documents rather than content. Neural networks removed the main constraint, the cost of production. When one firm can churn out thousands of "books" with no authors or editors, manually checking every item stops scaling.
What people who buy or publish content can take from this:
- Vet the supplier, not just the documents. A firm no one in the industry has heard of that suddenly delivers hundreds of titles is a warning sign.
- Read selectively. Even a quick review by an expert caught invented facts and fake illustrations. Four committees did not.
- Watch for volume anomalies. Thousands of titles from a new supplier on a compressed timeline is reason to ask questions.
- Don't rely on AI detectors alone. Here, what helped was a subject-matter expert and obvious factual errors.
For companies building products around text generation, this is also a reputational signal. Each scandal adds pressure on regulators, from mandatory labeling of AI content to requirements that suppliers prove authorship. Platforms that host third-party material, whether book marketplaces, educational services or knowledge bases, will face the same problem. Cheap volumes of generated text will get in through formal gates.
The Tamil Nadu story shows something unpleasant. The danger is not only model hallucinations. Fraud has become cheaper and larger in scale with their help, while control procedures have stayed the same. While the court waits for a report on 20 years of purchases, other organizations should check whether anyone actually reads what they buy.
- Nov 21, 2026Canva adds a "turn off AI" toggle and a filter that puts human-made works first in its library
- Nov 21, 2026Claude Code writes and publishes SEO articles three times a day on a schedule
- Nov 21, 2026An agent learned to cheat on its own review, and AI judges catch errors with under 55% accuracy
- Nov 21, 2026Wildberries cut its AI assistant's response time from 10.9 s to 3.6 s with DFlash 2
Key Questions Answered
How it came to light?
Perumal Murugan is the first Tamil author to be shortlisted for the International Booker Prize. He said that some of the books for the new library were written by AI. Imagine opening a reference book and finding invented dates, people who never existed, and pictures that could not exist in nature. That is what Murugan found: errors, made-up facts that look like neural network hallucinations, fake illustrations, and plainly fictitious author names.
What the authorities did?
The anti-corruption agency opened a criminal case and froze Agal Vilakku's accounts. The publisher's owner was arrested, and five officials were suspended from duty. The scandal triggered a review of procurement across the state. It uncovered more than 70,000 questionable books worth about 70 million rupees. The state court called the case exceptional and ordered the authorities to report on all book purchases over the past 20 years.
What this means for practitioners?
There is nothing technically sophisticated in this story. It required a cheap stream of plausible-looking text and a procurement system that looks at documents rather than content. Neural networks removed the main constraint, the cost of production. When one firm can churn out thousands of "books" with no authors or editors, manually checking every item stops scaling. What people who buy or publish content can take from this: Vet the supplier, not just the documents.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.