A rigorous, cross-jurisdictional legal examination of how copyright frameworks globally are straining under the rapid rise of generative AI.

Dr. Barry Scannell’s Thesis on AI and Copyright – Key Lessons and Reflections

by ChatGPT-4o

Dr. Barry Scannell’s doctoral thesis, A Critical Analysis of the Interaction Between Copyright Law and Artificial Intelligence, stands as a rigorous, cross-jurisdictional legal examination of how copyright frameworks globally are straining under the rapid rise of generative AI. This landmark work, rooted in comparative law, doctrinal analysis, and emerging case law, is both scholarly and deeply pragmatic. It provides critical insights for AI developers, regulators, and content and rights owners, advocating for systemic change to ensure legal frameworks keep pace with innovation while safeguarding human creativity.

🔍 Summary of Core Findings

Scannell structures his thesis around two primary dimensions:

1. AI Inputs – the use of copyrighted content for training purposes via Text and Data Mining (TDM);

2. AI Outputs – the legal (and philosophical) question of whether AI-generated works are eligible for copyright protection and if not, whether a new sui generis right should be introduced.

He analyzes these in the context of key jurisdictions (EU, US, UK, China, Australia, etc.), using case law, statutory frameworks, and policy developments (notably the EU AI Act and US fair use doctrine).

📌 Lessons for Stakeholders

🧠 For AI Makers

• TDM compliance must be embedded into model design: The thesis identifies copyright infringement risk where TDM is conducted without clear legal basis (e.g., EU’s Article 4 CDSM opt-outs, database rights). AI developers must ensure legal clarity around all training data sources.

• Sourcing transparency is critical: To gain user trust and regulatory legitimacy, developers should maintain auditable records of training data and comply with obligations like Article 53 of the EU AI Act.

• Fair use ≠ Free pass: Scannell warns that U.S. developers relying solely on fair use (e.g., OpenAI) face growing legal headwinds. Legal arguments based on transformativeness or public interest may not hold if memorization and substantial similarity are proven.

⚖️ For Regulators

• Harmonization is urgent: The current global patchwork of TDM exceptions (EU opt-outs, Japan/Singapore opt-ins, U.S. fair use, UK’s post-Brexit stasis) is untenable. Scannell urges supranational collaboration on minimum rights, licensing models, and enforcement regimes.

• Sui generis rights should be explored: For AI outputs, the thesis suggests that neither copyright inclusion nor exclusion is satisfactory. A sui generis system could recognize commercial investment in AI creation while preserving copyright's human-centric purpose.

• Clarify "general-purpose AI" obligations: The final EU AI Act version leaves uncertainty around whether foundational model developers (vs. downstream deployers) are responsible for TDM compliance. This risks regulatory arbitrage and must be addressed.

🏛️ For Content and Rights Owners

• Assert opt-out and license rights early: Scannell shows how publishers and creators must actively deploy TDM opt-outs (under Article 4 CDSM) and enforce licenses or terms-of-use to prevent unlicensed use of their works in training datasets.

• Litigation risk is rising – so is leverage: He documents how cases like NYT v. OpenAI and Getty v. Stability AItest the limits of fair use and copyright exceptions, suggesting rights holders may soon gain significant bargaining power.

• Human creativity must be preserved: Scannell positions copyright not only as an economic right, but as a cultural safeguard. Rights owners should frame licensing demands as supporting creative ecosystems, not just monetization.

🌟 Innovative, Valuable, Surprising and Controversial Contributions

✅ Innovative

• Dual Input/Output Framework: Scannell’s clean bifurcation of AI copyright issues into “input” (training data) and “output” (generated works) provides a lucid analytic scaffold. It helps rights holders and AI developers clearly distinguish legal obligations at each stage.

• Legal methodology + AI systems theory: His hybrid approach—doctrinal analysis supplemented by technical understanding (transformer architectures, tokenization, diffusion models)—is rare among legal scholars and improves cross-disciplinary credibility.

✅ Valuable

• Judicial Case Matrix: Chapter 3 offers perhaps the most comprehensive and comparative mapping of AI-related copyright cases to date—from Authors Guild v. Google to NYT v. OpenAI, including under-referenced EU and Asian cases. This is a go-to reference for legal teams.

• Sui Generis Proposal Details: Rather than just advocating sui generis rights, Scannell outlines scope, duration, and economic models—making his proposal actionable for policymakers and publishers alike.

🤯 Surprising

• Rehabilitating Turing via Lovelace: In a creative opening, Scannell situates the debate over machine creativity within the “Lovelace objection” and its rebuttal by Turing, offering a philosophical frame that both humanizes and historicizes AI’s legal challenges.

• Critique of EU AI Act’s Final Version: He takes a striking position that the AI Act—despite its ambition—may be ineffective on copyright due to vague provisions, unclear enforcement, and insufficient extraterritorial bite.

⚠️ Controversial

• No Copyright for AI Works—Ever: Scannell strongly defends the position that only humans can be authors under copyright law, with no room for revision. This will be seen as too rigid by some AI advocates and creators exploring hybrid authorship.

• Global Licensing Clearinghouse: He hints at the need for a universal, possibly WIPO-led, licensing body for TDM and AI training. While practical in theory, it raises geopolitical, technical, and competitive concerns likely to spark debate.

🧭 Conclusion: The Path Forward

Dr. Scannell’s thesis is a pivotal work for anyone operating at the intersection of law, AI, and creative content. It challenges both AI developers and rights owners to abandon outdated assumptions and prepare for a post-training, AI-integrated future where:

• Licensing becomes proactive, not reactive;

• Rights assertion tools are embedded in metadata and models;

• Regulatory frameworks reward transparency, not opacity;

• Sui generis models may become the norm for non-human works;

• International harmonization is no longer optional but essential.

If the AI era began with Ada Lovelace’s imagination, Scannell reminds us it will mature—or collapse—on our ability to protect that very human capacity.

📚 Bibliography

• Scannell, Barry. A Critical Analysis of the Interaction Between Copyright Law and Artificial Intelligence. PhD Thesis. University College Cork, 2024.

• EU AI Act 2024: https://data.europa.eu/eli/reg/2024/1689/oj

• LinkedIn post by Barry Scannell: https://www.linkedin.com/posts/ugcPost-7364609994320691202