Key Takeaways

  • The Madras High Court ordered an FIR and vigilance investigation into disputed public-library book purchases in Tamil Nadu.
  • Nearly 1,800 Tamil-language titles supplied by Agal Vilakku Pathippagam reportedly raised concerns about authorship, pricing, ISBN records, and factual accuracy.
  • The case highlights how AI-assisted publishing can expose gaps in procurement, attribution checks, and accountability across public institutions.

The Madras High Court has escalated concerns over allegedly fake, plagiarised, or AI-generated books acquired by public libraries in Tamil Nadu, ordering an FIR and an investigation by the Directorate of Vigilance and Anti-Corruption.

The court also directed authorities to freeze relevant bank accounts and withdraw suspect books for cataloguing and examination. The central issue is public procurement and oversight, rather than a confirmed AI security incident. It has not yet been established that artificial intelligence generated the disputed works.

According to the BBC, nearly 1,800 Tamil-language titles at Periyar Arivulagam Library in Coimbatore were reportedly supplied by Agal Vilakku Pathippagam. Reported warning signs included unfamiliar authors, inflated prices, questionable ISBNs, and books containing false information.

A preliminary review subsequently led to 523 books being removed from Chennai’s Connemara Public Library, while scrutiny expanded to library collections elsewhere in Tamil Nadu.

That sequence matters. Public libraries usually rely on publisher information, ISBN data, expert selection committees, vendor documentation, and established acquisition rules. Generative AI can make weak submissions look superficially credible, however. A manuscript can have a polished cover, plausible chapter structure, invented author biography, and confident prose while still containing fabricated claims or unattributed material.

Detecting synthetic text is not a reliable procurement strategy by itself. Detection systems can produce false positives, particularly when assessing translated material, formulaic reference works, or writing by non-native speakers. Human reviewers also cannot consistently identify AI-generated prose simply by reading it. The stronger control is provenance.

That means verifying whether an author exists, checking publisher and ISBN records, examining citations, reviewing sample chapters, documenting editorial responsibility, and retaining a traceable approval history. Price comparisons also matter. If a little-known title is priced unusually high, procurement teams have reason to request additional evidence before public funds are committed.

The court’s interpretation could influence those controls. MediaNama reported that the Madras High Court viewed AI-generated and plagiarised works as falling within “doubtful taste” under Tamil Nadu’s library-procurement rules. Such books should not be acquired without checks covering authenticity and quality, the court said.

Publishers including Agal Vilakku Pathippagam and Phoenix Publishing House are now part of a broader discussion about where responsibility sits when AI-assisted material enters institutional purchasing channels. Is disclosure by the publisher enough? Probably not. Libraries may also need contractual rights to request source files, evidence of editorial review, author identification, and details of any generative tools used in production.

The policy foundation already exists. India’s AI Governance Guidelines, launched in November 2025, emphasize fairness, accountability, transparency, India-specific risk assessment, and institutional preparedness. The Government of India has also introduced an AI Competency Framework for Government Officials, reflecting the growing expectation that public employees understand synthetic content, attribution risks, and documentation requirements.

Procurement workflows may need risk tiers. A general-interest novel and a medical reference book do not create identical harms if their content is false. High-risk categories could receive subject-matter review, while large or unusual orders could trigger publisher verification and conflict-of-interest checks.

Generative AI may reduce production time, but it does not transfer accountability away from authors, editors, distributors, or suppliers. Clear AI-use disclosures, defensible rights records, factual review, and auditable editorial processes can help distinguish legitimate AI-assisted publishing from bulk content operations.

The Tamil Nadu investigation will determine whether fraud, corruption, plagiarism, procurement failures, or other violations occurred. Whatever investigators conclude about the disputed books’ origins, the case has already exposed a practical weakness: institutions can purchase polished content faster than they can verify who created it, whether it is accurate, and why the public should pay for it.