European Union regulators have opened an investigation into Anthropic, the American artificial intelligence company behind the Claude family of AI models, focusing on the company's use of so-called "control directives" — the instructions and constraints baked into its AI systems.
According to Techzine Global, the probe centers on how Anthropic directs the behavior of its models through these built-in directives, raising questions about transparency, accountability, and who ultimately controls what an AI system will or won't do.
Control directives are essentially the rulebooks embedded inside AI models that shape how they respond — telling them what topics to avoid, how to handle sensitive requests, and what values to prioritize. While AI companies argue these guardrails are necessary for safety, regulators are increasingly asking whether such directives are disclosed clearly enough to users and businesses that deploy the technology.
The EU has become the world's most aggressive jurisdiction for technology regulation, having already enacted the AI Act, which classifies AI systems by risk level and imposes strict requirements on high-risk applications. An investigation into how a major AI developer programs behavioral constraints into its models fits squarely within that regulatory framework.
Anthropic, which has positioned itself as a safety-focused AI lab and has received billions in investment from major technology firms, now faces scrutiny over whether its internal safety mechanisms align with Europe's standards for transparency and user rights.
The case matters because it could set a precedent for how regulators worldwide treat the hidden instructions that shape AI behavior — rules that most users never see but that profoundly influence what these systems say and do.