EU Regulations & Policy

The EU AI Act's watermarking requirements for AI-generated content are fueling increased research into attack methods, following NTT DATA's December 2025 disclosure of vulnerabilities. Security experts predict an arms race between watermarking implementation and circumvention techniques as the regulation takes effect.

Leading AI researchers are building a compelling case that centralised, large-scale AI development is structurally unsafe and economically exclusionary — and their arguments carry direct implications for European regulators still defining what the AI Act means in practice. At stake is whether the EU will entrench the dominance of a handful of American and Chinese technology giants, or chart a genuinely alternative path.