Feature: AI
The potential penalties have a three-
The EU AI Act represents a seismic shift in how business is conducted, moving AI from a tech-team project to a board-level compliance and ethics priority
and medical devices. (Transport and specialised sectors often have their own specific compliance timelines, generally becoming applicable in August 2027.) The second condition is if the product
in question is required to undergo a third-party conformity assessment before it can be sold. This impacts: • Industrial machinery – AI used to control robotic arms, automated assembly lines or safety sensors in factories, falling under the Machinery Regulation.
• Electronics and radio equipment – AI integrated into smart devices, telecommunications hardware or industrial IoT sensors, under the Radio Equipment Directive.
• Medical devices – AI-powered diagnostic hardware or robotic surgery tools.
• Lifts and pressure equipment – AI used to manage safety protocols in elevators or industrial boilers. If a company manufactures smart
machinery or industrial IoT devices, it must follow both the AI Act and the relevant existing safety laws. This creates a double compliance burden since it must ensure the AI meets the AI Act’s data and transparency standards alongside the physical safety standards of the machinery or radio equipment directives, for example. For such products or systems, a CE Mark can’t be applied unless the high-risk requirements of the AI Act have been fulfilled.
Enforcing the AI Act Since the AI Act is a regulation and not a directive, it is directly applicable. This means that as of 2nd of February 2025,
the bans on prohibited AI became law across the entire EU automatically as no national law is required. If a company is not fully in compliance by December 2027, technically it is breaking EU law, regardless of whether its specific member state has implemented it. However, while the law exists, the
infrastructure to enforce it is not in place. Member states were supposed to designate National Competent Authorities and Market Surveillance Authorities by August 2025. Only a few member states have fully set up their oversight bodies, including Germany, which passed its KI-MIG draft bill in February 2026, despite the EU’s set deadline of August 2025. This means that if a member state hasn’t appointed a regulator, there is no one to report a violation to, and no clear mechanism for that country to issue fines. The European Commission is currently preparing infringement procedures against several member states for failing to set up these authorities on time. High-risk products often need a
third-party audit by a notified body, but currently there are almost no fully designated notified bodies specifically for AI. Without these auditors, companies can’t get the certifications they need to launch new high-risk AI products by the upcoming December 2027 deadline. For AI systems, like manufacturing
robots, fines are issued by national competent authorities, but because some member states have been slow to set these up, warning letters rather than fines might be issued in certain countries whilst they finalise their legal infrastructure.
tiered system based on the severity of the violation, according to whether that is considered unacceptable risk, non-compliance with obligations, or misleading authorities. The EU has introduced specific caps to ensure that a single fine doesn’t instantly bankrupt a small business.
Building a culture of AI trust The EU AI Act represents a seismic shift in how business is conducted, moving AI from a tech-team project to a board- level compliance and ethics priority. To truly succeed, businesses must bridge the gap between legal requirements and operational trust. To turn these regulatory hurdles into
a competitive advantage, organisations should adopt the core principles of the AI essentials framework: • Pillar 1 – Security and Data Integrity: Protecting against logic-based attacks and securing data pipelines.
• Pillar 2 – Transparency and Explainability: Ensuring stakeholders understand they are interacting with AI and can explain how decisions are made.
• Pillar 3 – Human-in-the-Loop and Accountability: Preventing irreversible AI decisions without human review.
• Pillar 4 – Fairness and Bias Mitigation: Ensuring AI does not discriminate or reinforce harmful stereotypes.
• Pillar 5 – Reliability and Hallucination Management: Using techniques like retrieval-augmented generation to ground the AI in factual data and mitigate fabrication.
• Pillar 6 – Societal, Ethical and Consequential Impact: Proactively assessing the broader effects of AI on the environment, workforce and society. It is vital to remember that AI-enabled
systems should not be built for a regulator but for end users. By the time the August 2028 transition windows close, the businesses that thrive will be those that didn’t just tick the box of the EU AI Act, but those that implemented AI essentials to create a transparent, resilient and human-centric AI ecosystem.
www.electronicsworld.co.uk July/August 2026 23
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44
