Last month in Brussels, LexisNexis and Teneo co-hosted “AI and IP: Defining the Boundaries of Ownership and Innovation,” a senior-level forum designed to turn fast moving policy into boardroom decisions.
Over two panels, policymakers, rights-holders and industry practitioners examined how Europe’s emerging framework will shape data transparency, licensing, authorship and AI-driven invention, and what corporate leadership teams should do next. Below outlines the key themes, regulatory shifts and practical actions for boards as Europe moves from AI policy design to implementation.
Transparency is the New Compliance Currency
Businesses cannot demonstrate lawful training of AI models, or responsible deployment of these models, without transparent and traceable data stewardship. But companies aren’t operating in a legal vacuum. Notably, the transparency provisions of the EU’s AI Act operate as market-entry controls, not just compliance challenges for companies. While the rules may now be in place, EU officials can be expected to develop a long tail of guidance and enforcement decisions, creating a catalogue of jurisprudence on how these rules will apply in practice.
In parallel, ongoing reviews of copyright and text and data mining rules will likely bring further adjustments as AI innovations develop at pace.
These themes were front and centre as the event opened with a panel titled “training your data on rightsholder material (Art. 50, AI Act),” moderated by Luca Bertuzzi, Senior Reporter at MLex, a specialist newsroom covering regulatory risks, with contributions from Alice Brancati, IP/IT Legal Advisor at Euroclear; Julien Chasserieau, Associate Director for AI and Data Policy at DIGITALEUROPE, a European digital-industry trade association; Stefano Gentile, Policy Officer – Copyright, European Commission; and Burak Özgen, Deputy General Manager of the European grouping of authors’ and composers’ societies (GESAC).
Just as companies have developed robust frameworks for financial reporting and corporate governance over time, a similar approach needs to be applied to transparency in AI, including evidence-grade data inventories, model bills of materials, training logs, opt-out capture and audit trails.
Without accepted standards and settled expectations, deployers hesitate and right-holders litigate. Maintaining best-in-class data governance as AI is deployed in organisations adds a level of protection from legal risks emerging from the use of AI tools. While standards and enforcement practices settle, actions that can be implemented in the meantime include sufficiently detailed training-data summaries, practical output-labelling methods and accepted rights-holder opt-out signals.
Copyright is Territorial, Your Technology Stack is Not
Copyright rules are still enforced country by country, but AI pipelines operate globally, often moving data across jurisdictions without friction. This mismatch creates cross-border enforcement and compliance risk, especially when training data or model outputs reach EU users.
Right-holders want opt-out instructions and licensing information to survive the entire AI development chain. That means rights signals must be machine-readable and persist from the moment content is ingested through training, fine-tuning and deployment. The European Commission and industry groups are exploring options such as embedded metadata, hashing and fingerprinting methods and structured opt-out registries. Sector needs vary widely, so no single technical solution will fit all contexts. However, companies can ensure that the AI models they implement adhere to a few basic principles, such as machine-readable licensing and opt-out signals and the use of fingerprinting technologies. Even when AI tools are developed, trained and implemented outside the EU, companies need to assume that they will need to adhere to EU laws when their AI is used in European markets.
When AI Contributes to an Invention, Humans Still Own the Pen
Copyright isn't the only field of intellectual property law where AI is challenging legal conventions. For patents across major jurisdictions, only a human can be named as an inventor. The unresolved issue is how much human contribution is enough when AI assists the inventive process. U.S. guidance sets out specific factors for assessing a “significant human contribution,” while recent German decisions take a somewhat more flexible view. At the European Patent Office (EPO), algorithms remain unpatentable “as such;” applicants must demonstrate a technical effect grounded in human-directed choices.
This came into focus in the “When AI innovates, who owns the patent?” panel, moderated by Inbar Preiss, Senior Reporter at MLex, with Angel Aledo, Chief Operating Officer and Chief Technology Officer at EPO; Daniel Friedlaender, Senior Vice President and Head of Brussels Office at CCIA Europe, the regional arm of the Computer and Communications Industry Association (CCIA); Nicola Halloran, Managing Director at Teneo; and Charles Bernard, Partner at Janson law firm.
Again, maintaining robust documentation in R&D, for prompts, model versions, datasets, human decisions and evidence of technical effect when AI is used as a tool in the innovation process, will reinforce the legal certainty for companies. In particular, when it comes to innovation projects involving multiple parties, it is important to define inventorship thresholds and ownership in the contract, not in court, and maintain evidence of how these contractual terms are met.
Checklist for Intellectual Property Guardrails in AI Deployments
- Map your models, covering training sources, licences, opt-outs respected and downstream restrictions.
- Label outputs, implement provenance and watermarking and put user-facing notices in products.
- Contract hard, flow down provider warranties, define inventorship and ownership in R&D and vendor agreements.
- Operationalise governance, name accountable owners, run model change control and log risk-relevant decisions.
- Prepare your evidence, assume an audit tomorrow, prove lawful data use and a significant human inventive step.
Field Notes for Leadership Teams
- Grace periods help providers but raise compliance risk for users. The AI Act gives developers extra time to meet obligations, but deployers still need reliable tools, documentation and rights information to prove lawful use. If providers deliver these late, the organisation running the model, not the provider, may carry the liability.
- Opt-out signals must survive your entire AI pipeline. Right-holders increasingly rely on machine-readable opt-outs (“do not train on this”). These signals often get stripped as data moves through crawling, cleaning and preparation. Reinforce them with hashing, watermarking and strong provenance so your systems automatically respect rights.
- Ownership in joint AI projects can become unclear very quickly. In EU-funded or multi-party work, agree roles early, record who contributed what and settle ownership before the project starts.
- The “person skilled in the art” standard is shifting. As LLMs spread, what counts as obvious or sufficiently described will change. Document the human creative input to protect your position.
- Europe grows faster when rules are simple. Clear regulation, access to capital and affordable energy are the practical conditions for scaling AI.
Teneo will continue to convene leaders as Europe turns transparency from a principle into an operating system for AI. If you would value a diagnostic or a board-level session to shape your organisation’s path forward, we are ready to assist.
Vision 2026: Where is the World Going in 2026 and Beyond?
At the same time, new findings from Teneo’s Vision 2026 survey of more than 350 public company CEOs and 400 institutional investors globally highlight how leaders are experiencing AI adoption on the ground. CEOs report that fewer than half of current AI projects are ROI-positive, with early gains concentrated in internal efficiency, administrative and customer-facing applications. Both CEOs and investors note that marketing and customer service use-cases show the greatest success, while AI strategies in security, legal and HR remain the most challenging due to their higher risk and complexity. As Ryan Cox, Global Head of Artificial Intelligence at Teneo, observes: “The first wave of AI returns came from easy efficiency wins. The next wave is about rewiring core processes that inevitably have a longer, bumpier ROI curve… These use cases are higher risk and have greater potential impact. You don’t rush them to market; you treat them as strategic change programs with board-level oversight, not experiments.” Notably, 86% of EU CEOs and 95% of EU investors believe the EU AI Act will have a positive impact on their businesses, underscoring the strong regional alignment between regulation and innovation.