Compensating theft (Why News Publishers Need Protection from AI), Pg8

Context:

• With generative AI models using vast amounts of online content, including journalism, without permission or compensation, the author argues for legal protections and equitable revenue-sharing mechanisms for news publishers.

Key Highlights:

• AI firms use news content as training data without seeking permission from publishers.
• This content scraping mirrors earlier patterns seen with Big Tech, where news media was undercompensated.
• The Department for Promotion of Industry and Internal Trade (DPIIT) has set up a copyright and AI committee, which is a timely initiative.
• News publishers need a legal right to opt in or out of data scraping and negotiate fair usage.
• There is a moral and economic imperative to ensure AI profits are shared with content creators.

Detailed Insights:

• AI systems like LLMs are trained using massive online content, including journalism, without consent or compensation to content creators.
• News publishers’ original work—produced with editorial rigour and expertise—is being used by AI models to generate monetised outputs.
• Earlier, Big Tech platforms (e.g., Google, Facebook) capitalised on news content for user engagement while giving minimal returns to publishers.
• The rise of AI-generated summaries and responses further reduces the need for users to visit the original news websites, weakening digital traffic and revenues.
• AI firms cite “fair use” to justify scraping and using copyrighted content for training models, a legally and morally contentious claim.
• There is no clear legal framework in India (or globally) defining rights of news publishers in the context of AI training data.
• Without intervention, this could lead to the economic erosion of journalism, impacting democracy, public awareness, and accountability.
• Publishers must be given a legal right to consent and license their data before it is used by AI developers.
• Governments should ensure news organisations receive fair compensation if their content powers AI systems.
• The DPIIT’s copyright and AI committee is a welcome move toward building a balanced regulatory framework.

Way Forward:

• Introduce clear copyright norms for AI training data.
• Mandate content licensing frameworks for news data use.
• Strengthen data consent protocols for publishers.
• Ensure equitable revenue sharing from AI outputs.

Scientific/Technical Concepts Involved:

• Large Language Models (LLMs): AI systems trained on large textual corpora to generate human-like responses.
• Web Scraping: Automated method of collecting data from websites, often done without permission.
• Fair Use Doctrine: A legal principle allowing limited use of copyrighted material without permission for purposes like education, commentary, or parody. Applicability to AI training is under debate.
• Generative AI Ethics: The debate on moral obligations of AI companies when using copyrighted or original content.