Archive · 25 Aug 2026
AI Copyright Debate Puts the Future of Information Ahead
Artificial intelligence systems learn from very large collections of information.
They do this by finding patterns between words, ideas and other data.
This is different from a search engine, which mainly finds information that already exists.
A Delhi court case involving ANI and OpenAI asked whether using copyrighted material to train AI can break copyright rules.
The case also raises questions about whether an AI answer copies a work or creates something new.
Copyright laws were mainly designed for copying by people, so courts may need to examine how machine learning fits those rules.
Paul Keller says AI could change how people get news, research and other knowledge.
He argues that public institutions should help build fair and transparent AI systems.
The article says protecting the whole information system may be as important as protecting individual copyrights.
The Delhi High Court’s ANI–OpenAI case raises questions about how copyright law should address AI training.
Generative AI learns statistical relationships from large datasets rather than simply retrieving or reproducing information.
Courts may need to distinguish between learning and reproduction, information and expression, and innovation and infringement.
Paul Keller argues that AI-generated summaries could reshape access to newspapers, journals, archives and other knowledge sources.
The article calls for public, transparent and accountable AI infrastructure and mechanisms that support the wider information ecosystem.
- Who
- ANI Media Pvt. Ltd., OpenAI OpCo LLC and the Delhi High Court are central to the discussion; Yuval Noah Harari and Paul Keller are also cited.
- What
- The article examines how copyright law should respond when generative AI trains on large amounts of information and produces new outputs.
- When
- The article refers to the Delhi High Court judgment and Paul Keller’s 2025 white paper, but gives no date for the judgment.
- Where
- The legal dispute was considered by the Delhi High Court in India.
- Why
- Generative AI uses computational learning and abstraction, creating uncertainty over whether training or generated responses amount to copyright infringement.
Copyright and Rights Protection
Open AI Access and Knowledge Commons
Use of copyrighted works in training
Copyright and Rights Protection
Copyright holders may argue that using expressive works to train AI could constitute reproduction or copying and therefore require legal protection or authorization.
Open AI Access and Knowledge Commons
AI developers may distinguish computational learning from ordinary reproduction, arguing that statistical abstraction produces new outputs rather than direct copies.
Focus of regulation
Copyright and Rights Protection
Copyright law can protect authors and other rights holders from unauthorized use of their works.
Open AI Access and Knowledge Commons
The article argues that copyright alone may overlook researchers, universities, libraries, archives and public institutions that sustain the wider knowledge commons.
Future information access
Copyright and Rights Protection
AI-generated summaries could reduce people’s direct engagement with newspapers, journals and archives, raising concerns about control of information.
Open AI Access and Knowledge Commons
Publicly governed AI infrastructure could provide transparent, accountable and equitable access to knowledge while supporting a sustainable information ecosystem.
- Case
- ANI Media Pvt. Ltd. versus OpenAI OpCo LLC
- Court
- Delhi High Court
- Technology
- Generative AI and large language models
- Core legal question
- Whether using copyrighted works in AI training constitutes reproduction or copying
- Key distinction
- The article contrasts information with expression and learning with reproduction
- Referenced paper
- Paul Keller’s 2025 white paper, “Beyond AI and Copyright: funding a sustainable information ecosystem”
- Proposed response
- Public AI infrastructure that is transparent, accountable and equitable