Fondamentaux du Droit d'Auteur et de la Propriété Intellectuelle pour l'IA Générative

Ce qu'elle couvre

Cette formation couvre l'ensemble du cycle de vie de la propriété intellectuelle dans l'IA générative : légalité de l'ingestion des données d'entraînement, droits sur les sorties des modèles, clauses d'indemnisation, normes de tatouage numérique et jurisprudences récentes en Europe et aux États-Unis. Les participants travaillent sur des scénarios concrets impliquant des LLM, des générateurs d'images et des assistants de code afin d'identifier les risques et les stratégies de mitigation. La formation alterne exposés d'experts, études de cas et exercices de rédaction de politiques. Les équipes repartent avec un cadre opérationnel pour évaluer le risque PI dans tout flux de travail assisté par IA.

À l'issue, vous saurez

• Assess whether a given AI training dataset use case falls within EU TDM exemptions or US fair use doctrine
• Identify clauses in AI vendor contracts that transfer, limit, or indemnify IP liability, and flag red flags for legal review
• Determine who — if anyone — holds copyright in a specific AI-generated output under current EU and US law
• Design a basic internal AI content policy covering attribution, disclosure, and acceptable use aligned with current regulations
• Explain the role of C2PA watermarking and content credentials in proving provenance and reducing infringement risk

Sujets abordés

• Training data provenance and fair use / text-and-data mining (TDM) exemptions
• Ownership of AI-generated outputs under EU and US copyright law
• Indemnification clauses in AI vendor contracts (OpenAI, Adobe, Microsoft Copilot, etc.)
• Digital watermarking and content provenance standards (C2PA)
• Recent EU and US case law: Getty Images v. Stability AI, NYT v. OpenAI, Thaler v. Vidal
• EU AI Act obligations intersecting with IP and copyright
• Open-source model licensing (Apache 2.0, CC-BY, RAIL licences)
• Internal policy drafting: acceptable use, attribution, and disclosure requirements

Modalité

Delivered as a one-day in-person or live-virtual workshop split roughly 50% expert instruction and 50% case study exercises and policy drafting. Materials include a case law digest (updated quarterly), contract clause annotation worksheets, and a reusable IP risk assessment checklist. Remote delivery uses breakout rooms for group exercises. An optional 90-minute follow-up Q&A session can be scheduled 2-4 weeks post-workshop to address implementation questions.

Ce qui fait que ça marche

• Involve legal counsel alongside content and product teams in the same session to align understanding and accelerate policy decisions
• Anchor learning in real vendor contracts and real case law rather than hypothetical examples to drive immediate applicability
• Assign a policy owner before the training ends so that the IP risk checklist produced in the workshop has a clear next action
• Schedule quarterly refreshers or update briefings given the rapid pace of litigation and regulation in this space

Erreurs fréquentes

• Assuming that outputs from paid commercial AI tiers are automatically IP-safe and indemnified without reading the vendor contract carefully
• Treating AI-generated content as fully copyrightable company IP without verifying human authorship thresholds required by national law
• Ignoring training data provenance when fine-tuning open-source models on proprietary or scraped datasets
• Conflating EU AI Act compliance with copyright compliance — they are separate regulatory regimes with different obligations

Fournisseurs à considérer

• Fieldfisher AI Legal Training (EU)www.fieldfisher.com/en/services/technology/artificial-intelligence →
• Bird & Bird AI Academywww.twobirds.com/en/capabilities/artificial-intelligence →
• WIPO Academy – AI and IPwww.wipo.int/academy/en/ →
• CMS Law – AI Training Workshopscms.law/en/int/service/artificial-intelligence →

Sources

• EU AI Act – Copyright and TDM provisions (Article 53) →
• US Copyright Office – Copyright and Artificial Intelligence (2024 Report) →