New Machinery Regulation enters into force
Physical AI: conflicting priorities of regulation and dynamic innovation
Three questions for Walter Reithmaier, CEO Division Industry & Infrastructure, TÜV SÜD.
How does TÜV SÜD align the conformity assessment required by the Machinery Regulation with the concurrent requirements of the new EU AI Act and the EU Cyber Resilience Act? What options do corporate managers have here?
Reithmaier: The Machinery Regulation, the EU AI Act, and the Cyber Resilience Act address different protection objectives but have some overlaps in content, thus requiring a holistic approach when dealing with digital, networked, and AI-based machinery.
Therefore, companies need to adopt a risk-based and interdisciplinary approach at an early stage and consider compliance implications right from the start of development. The first step is to identify any relevant hazards along with AI and cyber risks, determine appropriate countermeasures, and to align them with regulatory requirements as well as the corresponding conformity assessment procedures.
So, managers should make sure that safety, security, and AI governance are integrated in a single, unified approach rather than being handled by separate organizational units. Taking regulatory requirements into account early on reduces risks, avoids costly rework, and accelerates market introduction of new products.
TÜV SÜD helps manufacturers and operators navigate the entire compliance process: from consulting and technical guidance to appropriate assessment processes and the use of digital tools to ensure efficient and robust proof of compliance.
Conformity assessment of self-learning or AI-based machines is expected to be particularly complex or even impossible as machines such as humanoids are capable of performing a huge variety of tasks. Is there a solution?
Reithmaier: Conformity assessments of self-learning or AI-based machinery present greater challenges, but remain manageable. The known hazards are the same. What’s new is that AI can influence safety-related decisions or augment traditional control systems. This requires joint assessments for market introduction and operation, i.e. a thorough risk assessment, suitable technical measures, an assessment as per the Machinery Regulation, the AI Act, and applicable standards, as well as ongoing monitoring of safety-related decisions during operation.
General standards and generic solutions alone are insufficient here. It requires domain-specific approaches to solutions and evaluations such as those for automated and deterministic detection of hazardous situations. These should then be tested in collaboration with industry partners and within the framework of pilot projects in order to develop viable standards as well as certification and monitoring concepts for their market introduction and operation.
I firmly believe that regulation can keep up with the progress of technology. What it takes is close collaboration between industry, research, authorities, and independent testing organizations.
Do you consider existing regulation a hindrance to innovation in Physical AI that keeps European companies from performing well in this promising market? Or could standards and certifications even help scale new technologies more quickly?
Reithmaier: Regulation and innovation are often considered irreconcilable opposites. I don’t see it that way. Good regulation builds trust, which is a key prerequisite for widespread acceptance of new technologies.
Therefore, regulation of Physical AI should not be considered a hindrance to innovation, but a necessary safeguard providing safety, guidance, and prediction of liability risks. Since technological developments frequently outpace the development of existing standards, a pragmatic, risk-based approach is required. Standards and certifications provide guidance to companies but should be applied with some flexibility so as not to impair the development of new solutions. It is crucial that regulatory protection objectives are met through fact-based risk assessments, robust proof, and suitable conformity assessment procedures.
More on the Machinery Regulation
Jens Müller, zertifizierter Sachverständiger für Maschinensicherheit
