
Tech • IA • Crypto
Artificial intelligence is reshaping legal frameworks, but current rules largely extend existing laws while leaving major questions about ownership and responsibility unresolved.
The regulation of AI builds on existing laws rather than replacing them. Frameworks such as the GDPR and national legal systems still apply, while the EU AI Act adds a new layer focused on risk classification. Systems are categorized from minimal to “unacceptable” risk, with strict bans on practices like social scoring or manipulation of vulnerable users.
Different regions are taking distinct approaches. China introduced strict rules on generative AI as early as 2023, while the United States regulates more incrementally. The European Union aims for a balance between innovation and protection, though it faces criticism for potentially slowing technological progress.
The EU framework emphasizes transparency obligations, including labeling AI-generated content such as deepfakes. It also imposes compliance duties on platforms and developers, particularly for high-risk systems. Some provisions have been delayed to give companies more time to adapt and support technological competitiveness.
Policymakers are prioritizing safeguards against harmful uses, especially deepfake content and risks to minors. New measures include stricter controls on sexually explicit synthetic media and enhanced scrutiny of AI systems that could exploit psychological vulnerabilities.
The legal environment remains fluid. The rapid evolution of AI—especially generative and agent-based systems—means laws cannot remain static. Regulations are expected to evolve continuously, with courts playing a key role in interpreting unclear areas.
Content generated entirely by AI is generally not protected by copyright under current interpretations in both the EU and the US. A simple prompt, even if sophisticated, is not considered sufficient human creativity to grant ownership rights.
When AI-generated material is significantly modified by a human, copyright protection may apply. However, creators must demonstrate and document their creative input, such as edits or transformations, to qualify for legal protection.
If AI outputs are based on proprietary inputs—such as personal datasets, voices, or images—rights may still apply. Ownership can depend on whether the user holds rights over the input material and how it is used within the system.
AI-generated software code is typically not copyright-protected if produced autonomously. Additionally, “contamination” risks arise when generated code incorporates open-source components, potentially imposing licensing obligations on otherwise proprietary products.
Liability is not yet fully settled and may involve users, developers, or platforms depending on the case. Under the Digital Services Act, platforms must monitor and mitigate harmful content, effectively taking on partial regulatory responsibility.
Courts are expected to handle increasing disputes over authorship, liability, and misuse. Current ambiguities may lead to situations where AI-generated works are effectively owned by no one, challenging traditional intellectual property systems.
Legal practice is already evolving, with professionals using AI tools while navigating strict confidentiality rules. Routine advisory work is increasingly automated, pushing lawyers toward litigation, negotiation, and complex advisory roles.
AI is not replacing existing legal systems but exposing their limits, forcing continuous adaptation as lawmakers, courts, and industries redefine ownership, responsibility, and acceptable use.