LA366_Lecture_3_LO

Lecture Overview

• Course: WARWICK LA366: Lecture 3 - Copyright Part II

• Instructor: Dr. Luminita Olteanu

Key Topics

Part 1: Copyright and Access to Knowledge & Information Flows

• News aggregators

• Press publishers' rights

• Access to educational materials

Part 2: Copyright and Generative AI

• Definition of Generative AI

• Training GenAI - text and data mining issues

• Copyright issues with AI-generated works

Copyright in News

• Expression vs. Facts:

  • Copyright protects specific expressions of news (e.g., text of an article) but not the underlying facts.

  • Different agencies can express the same facts in various ways.

  • Challenges for news aggregators regarding copyright when reproducing headlines and snippets.

  • Example: "Company X acquires Company Y for £Z million and promises not to fire employees."

Protection under the Berne Convention

Headlines and Snippets

• Historical Context:

  • Proposal to amend the Berne Convention to protect distinctive titles failed, particularly due to practical issues raised by the British delegation.

• Article 2 does not specify a minimum length requirement for copyright protection.

• Member countries have flexibility in their protection of short phrases based on national laws.

Infopaq International A/S v Danske Dagblades Forening [2009]

• A case concerning the copyright of headlines and snippets:

  • Infopaq operated a media monitoring service that required copyright authorization.

  • Dispute over whether storing 11-word excerpts from articles was permissible.

  • CJEU concluded protection applies if reproduced elements express intellectual creation.

Copyright in the US

• Titles and Short Phrases:

  • Generally uncopyrightable due to insufficient authorship.

  • Example: Ninth Circuit held specific lyrics could be original in a lawsuit, challenging broad views of copyrightability.

Fairfax Media Publications v Reed International Books [2010]

• Important questions on whether copyright extends to headlines:

  • Most headlines in question deemed short factual statements with little literary merit.

Quotation Rights under Article 10

• Permits quotations from lawfully published works under fair practice, contingent upon:

  • Purpose of use

  • Size of quotation

  • Nature of the original work

  • Impact on the market for the source work.

Press Publishers' Rights

Background

• Tensions between news organizations and digital platforms led to the EU's adoption of a new press publishers' right in 2019 aimed at supporting press sustainability.

• Opposition centered around restricting online information sharing.

Types of Press Publishers

• Classification by Ownership:

  • Commercial/Private

  • Non-profit organizations

  • Public/state-related

  • Mixed/hybrid models.

2019 CDSM Directive

• Defines press publications with rights expiring two years post-publication; includes provisions for revenue share with authors.

• Various exclusions from rights and protections.

National Implementations of Revenue Sharing

• Different countries’ adaptations to revenue sharing rules with varying compensation percentages for authors.

Digital Content Bargaining Frameworks

Australia

• News Media and Digital Platforms Mandatory Bargaining Code designed to address power imbalances in negotiations between news media and platforms like Google and Facebook.

Canada

• The Online News Act emphasizes fairness in the digital news marketplace and outlines negotiation rights regarding news content.

Access to Educational Materials

• Challenges persist with 75% of academic research being paywalled, affecting accessibility for students and researchers.

Open Access Movement

• Berlin Declaration promotes wide access to knowledge and open access publishing initiatives to combat high costs.

• Costs involved in publishing open access can also be a barrier.

Generative AI

• Definition: AI that can create outputs, such as text or images, requiring considerable training data.

• Hallucination: Can produce plausible but inaccurate responses.

Interaction of Copyright and Generative AI

• AI requires significant quantities of data, raising copyright issues when training on copyrighted materials.

• Lack of legal protections for creators of copyrighted works used in AI training.

Arguments for and against TDM (Text and Data Mining)

Arguments for TDM Without Authorization

• Reproduction is not for enjoyment; it serves as a tool for information extraction.

• Restrictions create barriers for research and development.

Counter Arguments

• Generative AI should not exploit human creativity without compensation.

• Exempting TDM from copyright risks the reduction of artistic production to mere inputs.

Global Perspectives on Generative AI and Copyright

Australia

• Current law lacks broad exceptions for TDM, relying on fair dealing provisions.

EU and Other Countries

• Variability across jurisdictions concerning TDM rights, emphasizing balancing fair use and copyright protection.

AI as a New Category of Copyrightable Works

• Under certain conditions, AI-generated outputs may qualify as works under copyright law if they reflect human input or creativity.

Conclusion: Political Economy of Copyright and AI

• Questions arise regarding the need for incentives for human creativity when AI technologies can produce works at scale and lower cost, prompting discussions on alternative remuneration mechanisms.