Copyright lecture 3

US Copyright – Fair Use: What are the four factors to consider?

• (PANE) ezelsbruggetje

• Purpose & character – non-commercial, criticism, comment, teaching, research favored.

• Nature of work – factual > creative.

• Amount & substantiality – how much and how important.

• Effect on market/value – does it harm the original’s market?

EU Copyright & AI Training

• TDM (Text & Data Mining): automatic analysis of data.

  • Art. 3 CDSM (Digital Single Market Directive) : research only.

  • Art. 4 CDSM (Digital Single Market Directive): commercial use allowed unless rightsholders opt-out.

  • Art. 17(7) CDSM: Protects user rights on platforms (parody, review, quotation). Not a training exemption.

• EU AI Act Art. 53:

  • 53(c): Have a copyright policy and follow Art. 4 CDSM

  • 53(d): Publicly list what content was used for training

How AI Training Works

• Web crawling: collects HTML, follows links; respects robots.txt.

• Web scraping: downloads files directly (riskier).

• Alt text: Used by crawlers to “label” media.

Machine-Readable Opt-Out

• For:

  • Empowers creators to signal “do not train on my content.”

  • Could force companies to negotiate licensing deals.

• Against:

  • Only big players (Google, OpenAI) can realistically comply at scale.

  • May strengthen dominant firms, harm small competitors.

Likeness Laws

• USA:

  • NO FAKES Act (draft): Prohibits unauthorized AI voice/image replication.

  • Tennessee ELVIS Act (2024): First state law protecting image/voice likeness.

• EU:

  • No EU-wide law; national personality rights apply (Germany, France stricter).

What did the Beijing Internet Court decide on AI-generated works?

It granted copyright to an AI-generated image, saying it showed enough human intellectual input and originality.

What is the EU Commission’s 4-step test for copyright?

(HICE) ezelsbruggetje

1. In a creative domain (literary, science, art)

2. Human effort

3. Creative choices

4. Expression of intention

What is a “work” protected under EU copyright law?

No single EU directive defines it.

• Berne Convention Art. 2 → any production in the literary, scientific or artistic domain, in any form.

• A work = “the author’s own intellectual creation”.

Who is considered an “author” under EU copyright law?

The author is a natural person (human) who makes free and creative choices, giving the work their personal touch. Machines/tools may assist, but cannot be the author.

Test for AI-assisted output: result of human creativity?

• Domain: literary, scientific, or artistic

• Human intellectual effort

• Originality/Creativity: author’s own + free & creative choices

• Expression: creativity must be expressed

Key points on authorship under EU copyright (incl. AI-assisted works)

• EU law: little rules; national law dominates

• Only creatively active persons = authors

• Joint authors = inseparable contributions + concerted effort

• AI-assisted works: usually person indicated as author

What are some creative uses of AI in music?

VAIT ezelsbruggetje

• Audio production (EQ, mastering, effects)

• Individual instruments (drums, bass)

• Voice cloning

• Text-to-music generators

What are the purposes of copyright?

RSS ezelsbruggetje

• Reward & respect the creator

• Stimulate innovation

• Stimulate cultural production

  • Gives creators a monopoly so they’ll invest in creation

Municipal Court of Prague on AI-created works?

AI-created work ≠ personal creation

The false Gen-Ai dichotomy

1. Het gebruik van GenAI betekent niet automatisch wie de auteur is.

2. Er is geen betrouwbare manier om precies te bepalen welk deel copyright krijgt.

3. Verschillen tussen landen kunnen internationale auteursrechten lastig maken.

Ramifications of differing copyright rules for international markets?

NRCC

• No uniform authorship rules across countries

• Regulatory uncertainty weakens copyright’s market function

• Creators may venue shop for better protection

• Can distort markets and competition

Two tricky prerequisites for a coherent international copyright market?

• Common, sufficiently specific standard for evaluating works → needs broad international harmonization

• Detailed understanding of how each work was developed → is this even possible?

Why is it hard to draw the line for GAI-created works?

GQCS

• GAI tools & user skills vary

• Quantitative rules aren’t enough

• Case-by-case checks are hard to scale

• Some parts may be public domain in certain countries

What are the two main methods for handling AI in copyright?

1. Tracking – monitoring AI use and contributions

2. Trust (current) – relying on users/creators to accurately report contributions

Creativity tracking with watermarks – problems?

• Visible, invisible, or metadata watermarks can usually be removed

• Few robust watermarks exist (Prof. Ben Zhao)

• Not recommended by EU & US briefings

• Cannot detect GAI use or which elements were created

Prerequisites for a functional watermarking system?

• Significant technological innovation

• Harmonized international standards & enforcement

• International policing of non-compliant tools

• Comprehensive tracking of creative process

• Essentially: major reengineering of international copyright

Two options for handling AI-created works in copyright?

• Option 1 – Difficult Status Quo: Enforce rights based on GAI use → hard to audit, doesn’t show creative contribution

• Option 2 – Redesign: Accept new creative modes & address copyright challenges directly

Challenges of new creative production realities?

• Substantial similarity tests may need reinterpretation

• Question whether copyright still works as market incentive/reward for creators

Incentives in the new creative economy?

• Creators gain value from their overall brand, not individual works

• Platforms reward frequent content creation → more visibility & engagement

• GAI tools help meet this constant content demand

• “Copy-mining” favors those with most computational power, giving them rent-seeking control over others

Risks of a world without copyright?

• Production may shift to low-cost, low-effort works

• Could reduce creation of high-investment, high-quality works

• Market might favor automatic/functional art over truly creative works

Key features of the future of creativity?

• No temporal boundaries

• No spatial boundaries

• No content boundaries