Understanding the EU AI Act for Machine-Builders
The EU AI Act aims to regulate artificial intelligence systems within the European Union, establishing a framework that impacts machine-builders significantly. This legislation promotes the development and use of human-centric and trustworthy AI, which is crucial as AI technology increasingly influences both personal and professional environments.
Risk Levels Defined by the EU AI Act
The Act categorizes AI systems into four levels of risk, with the highest risks being associated with systems that could adversely affect health, safety, or fundamental rights. For machine-builders, understanding these risk levels is essential, particularly as they may integrate AI into control systems for automation and robotics.
Compliance Timeline
Key dates for compliance are critical for machine-builders:
- 1 August 2024: The Artificial Intelligence Regulation (EU) 2024/1689 comes into force.
- 2 August 2026: Most regulations will apply.
- 2 August 2027: High-risk AI systems used as safety components must comply with the regulations.
Machine-builders supplying machinery to the EU must comply with the AI Act, even if the systems are operated outside the EU but impact users within it.
Compliance Steps for Machine-Builders
Before an AI system can enter the market, it must be CE-marked. This requirement applies whether the AI system is sold independently or embedded within other products:
- If embedded, the product manufacturer assumes the role of the provider and is responsible for compliance.
- The AI Act differentiates between providers (e.g., machine-builders) and deployers (e.g., end-users).
Conformity Assessment Procedures
Similar to the Machinery Directive, the AI Act outlines conformity assessment procedures, which may involve:
- Self-certification for many high-risk AI systems.
- Utilizing Notified Bodies for third-party assessments, which require re-certification every five years.
Complying with harmonised standards will facilitate adherence to the requirements, although these standards are still being developed.
Declaration of Conformity
Once compliance is achieved, a Declaration of Conformity (DoC) can be issued. If the AI system is part of another product, the DoC may be included in that product’s documentation.
Post-Market Obligations
After market entry, providers must conduct ongoing monitoring of high-risk AI systems throughout their lifecycle. Obligations include:
- Reporting any serious incidents to the appropriate authorities.
- Implementing automatic event logging for better surveillance.
Substantial modifications to AI systems post-approval necessitate re-assessment for compliance.
Penalties for Non-Compliance
The AI Act stipulates strict penalties for non-compliance, which can include fines of up to €15 million or 3% of the organization’s annual global revenue. Higher fines apply to prohibited AI systems.
Understanding the EU AI Act is essential for machine-builders, as it not only outlines the necessary steps for compliance but also aims to foster innovation and investment in AI technologies. By adhering to this legislation, machine-builders can ensure their products are safe, effective, and aligned with ethical standards.
