EU AI Act transparency rules take effect for firms

The European Union's AI Act has reached a key implementation phase, with Article 50 transparency rules now in force for organisations building and deploying AI systems.

The measures apply to businesses operating in, or selling into, EU and UK markets where they fall within the scope of the law.

Under Article 50, organisations must disclose when users are interacting with an AI system. They must also label certain AI-generated or AI-manipulated content and provide information about the use of biometric and emotion-recognition tools.

These transparency provisions sit alongside wider obligations on risk management, data governance and technical documentation for systems covered by the Act's risk-based regime. Although some high-risk requirements have longer transition periods, legal and compliance specialists say this milestone marks a shift from preparation to enforcement for many companies.

Industry figures say the new rules respond to growing concern over opaque AI use in consumer and enterprise services. They also warn that regulation and organisational practice remain out of step with the pace of deployment.

Transparency milestone

RAIDS AI, which focuses on monitoring AI systems in production, described the new transparency phase as necessary but incomplete.

"The next EU AI Act milestone is an important step forward, but it also highlights an enduring problem: AI regulation is moving far more slowly than AI itself.

"From 2 August 2026, organisations will need to meet new transparency obligations, including making clear when people are interacting with AI systems or when content has been generated or manipulated by AI. That is a positive and necessary development. People have a right to know when AI is being used, particularly as the technology becomes more embedded in the services and content they encounter every day.

"But transparency alone is not enough. The bigger question is how we ensure that AI systems are safe, reliable and accountable once they are deployed in the real world. That is why the delay to the implementation of rules for high-risk AI systems, following the EU AI Act Omnibus announced earlier this year, is significant.

"High-risk AI systems fall into two categories. The first is AI systems that are themselves products, or safety components of products, such as medical devices, machinery, vehicles, aviation, lifts, toys or other regulated products. The second covers AI systems that can significantly affect people's health, safety or fundamental rights in specific use cases listed in the AI Act.

"In short, these systems are affecting people's opportunities, finances, health, wellbeing and access to essential services, yet the timetable for mandatory compliance has been pushed further into the future. The danger is that organisations interpret this as permission to wait. But they cannot afford to do that. AI systems are already being used at scale, and risks such as bias, hallucinations, model drift, security vulnerabilities and unintended behaviour do not wait for regulatory deadlines.

"We need to stop treating AI safety as something that begins when the law catches up. Regulation will always lag behind technology to some extent, but responsible organisations should not. Businesses have an ethical imperative to take AI seriously. Those that are doing so are putting the right safeguards, monitoring and governance in place now.

"The challenge is not to choose between innovation and regulation. We need both. AI has enormous potential to improve productivity, decision-making and services, but that progress will only be sustainable if people can trust the systems being deployed. Transparency is a start, but continuous oversight that monitors AI behaviour in production, after deployment, is what will determine whether AI can be trusted at scale."

Nik Kairinos, Chief Executive Officer and Co-founder of RAIDS AI, said the delay to high-risk AI rules could encourage some organisations to postpone action even as systems are already being deployed at scale.

The EU has staggered deadlines for obligations relating to systems classified as high risk. Compliance timetables run into 2027 for some products and use cases, raising concern that organisations may delay investment in controls until closer to those dates.

Innovation concerns

Health AI founder Jiva.ai raised questions about the broader design of the regulatory regime. The company operates in clinical and life sciences settings that already face strict sector regulation.

Dr Manish Patel, Chief Executive Officer and Co-founder of Jiva.ai, said: "The EU AI Act tackles vital safety concerns, particularly around transparency, fundamental rights, accountability and systemic risk. Yet we have to ask: does adding a broad horizontal AI compliance layer help innovation or simply slow it?

"Stringent bureaucracy carries real innovation risks, threatening to delay deployment in regulated sectors such as life sciences, where firms already face medical, data protection, clinical, cyber and procurement regulation. These heavy documentation and compliance burdens will fall disproportionately on smaller firms, and they may become obsolete as AI architectures rapidly evolve.

"The UK must maintain EU compatibility for market access but domestically pursue an adaptive, outcome-based framework that uses sandboxes and sector-led expertise to make the UK the global hub for responsible and efficient innovation."

UK-based companies that export AI systems or services into the EU fall within the scope of the Act, despite the UK's looser, principles-based approach. Policy advisers say this creates a dual-track compliance picture for British firms seeking EU access while also responding to evolving domestic guidance.

Competitiveness test

Investors are watching how Europe's approach influences capital flows and start-up formation in AI, particularly compared with the US and parts of Asia. Some see benefits in the legal clarity that follows several years of legislative debate.

Jeremy Brown, Investment Principal at Anthemis, said: "With the final provisions for the EU AI Act now in place, the EU faces a global test to prove that a highly regulated AI market can remain globally competitive.

"The Act finally provides businesses with greater regulatory certainty. Rather than building against a backdrop of evolving rules, organisations now have a clearer framework to invest, innovate and scale responsibly.

"Compared with markets like the US, Europe has taken a more rules-based approach to AI. Much like how the EU set the global benchmark for data privacy through GDPR, the question now is whether it can do the same for AI without sacrificing competitiveness.

"The success of the AI Act will not be judged by how many businesses comply with it, but by whether Europe continues to attract the world's best talent, ambitious AI companies and long-term investment. If it does not, pressure to refine the framework will inevitably grow."