The headlines are constant, and the looming deadlines of the EU AI Act are no longer a distant concern. For many organizations, the transition from understanding the legal text to implementing technical solutions feels like staring across a vast, foggy canyon. The legislation defines what must be achieved, but it is notoriously silent on how to actually achieve it.
At the Data Innovation Summit 2025 in Stockholm, Patrick Ridell from the Swedish Institute for Standards provided the missing link. The session, From Policy to Practice: Bridge Compliance for AI Act through European Standards, serves as a strategic roadmap for leaders, developers, and compliance officers navigating the maze of European regulation.
For any organization aiming to move AI systems from experimental “black boxes” to certified, market-ready products, this presentation offers indispensable clarity.
The “How-To” Manual for EU Law
The EU AI Act is more than just a standalone set of rules; it is a critical component of the New Legislative Framework. While the title of the talk suggests something modern, this framework has governed European product safety for nearly two decades. It establishes the ecosystem for CE marking, conformity assessments, and the market surveillance required to keep products in circulation.
However, a law is a high-level document. It utilizes broad terms like “transparency”, “robustness”, and “human oversight.” For a developer sitting at a terminal, these concepts can feel frustratingly abstract. As Patrick noted during the summit:
“The legal texts can be quite seemingly ambiguous and abstract… maybe not your favorite Friday reading. But that’s where we will have the standards to actually operationalize the requirements.”
The core takeaway is simple: The Act provides the “What”, but the Standards provide the “How”.
High-Risk Systems: The Core of the Challenge
Much of the industry anxiety surrounding the AI Act focuses on High-Risk AI systems—those integrated into critical infrastructure, biometric identification, or credit scoring. These systems are subject to “essential requirements” that form the basis of the new regulatory regime.
The session breaks down the pillars that these standards will eventually codify:
- Quality Management Systems (QMS): The operational heart of the organization.
- Risk Management: A continuous, iterative process of identification, analysis, and mitigation throughout the AI lifecycle.
- Data Governance: Strategies to ensure training and test data are of high quality and free from prohibited biases.
- Logging and Traceability: Creating a reliable “paper trail” to explain why a system reached a specific conclusion.
- Human Oversight: Defining the protocols for human-in-the-loop intervention and system resilience.
The Secret Weapon: Presumption of Conformity
Perhaps the most significant insight for businesses is the role of Harmonized European Standards.
Why should a startup or an enterprise prioritize a voluntary standard? The answer lies in legal certainty. When an organization follows a harmonized standard officially cited by the European Commission, it benefits from a “presumption of conformity”. Essentially, being certified against these standards provides a legal guarantee of compliance with the AI Act. This serves as the ultimate safeguard against hefty financial penalties which can reach up to €35 million or 7% of total global annual turnover.
The Risk of Waiting
While the general application for high-risk requirements does not take full effect until August 2026, waiting until the final hour is a risky strategy. Through a case study of a fictitious startup, “Alpha Tech,” the presentation demonstrates how early engagement with national committees allows companies to:
- Access draft standards before they are finalized and made public.
- Integrate AI-focused controls into the development lifecycle today.
- Avoid the inevitable bottleneck of certification bodies that will occur once the law is fully active.
Organizations that act now will not be struggling with compliance in 2026 and they will be leading the market in innovation.
Why the Full Session is Essential Viewing
This talk represents a rare connection between the legal and technical worlds. It successfully translates a complex regulatory burden into a structured, achievable plan for business growth and ethical AI deployment.
Hyperight members receive exclusive access to the full video of this session, which includes:
- A detailed breakdown of the AI Act’s implementation timeline and the “gradual phase-in” of prohibitions.
- Insights into how companies can join national committees to help shape the standards themselves.
- The “IKEA instructions” analogy for understanding EU law.
- Practical methodologies for aligning legal, technical, and policy teams within a single organization.
Become a Member to Watch the Full Session
Looking Ahead: Data Innovation Summit 2026
As the industry moves toward a future of regulated, trustworthy AI, these conversations are more vital than ever. The dialogue continues at the 11th Edition of the Data Innovation Summit. Join the global data community in Stockholm to explore the next frontier of AI, governance, and technical excellence. Stay tuned for further updates on what promises to be the most significant gathering yet.
