Recently, the European Parliament voted and passed the landmark EU AI Act. It's the first of its kind and sets a benchmark for future AI regulations worldwide.
The EU AI Act lays the foundation for AI governance, and it's pertinent for organizations delving into AI systems to comply with the legislation, build robust and secure AI systems, and avoid non-compliance fines.
 |
| Photo by Karolina Grabowska via Pexels |
My three key takeaways from the legislation are as follows:
- The Act introduces the definition of an AI system:
- "An AI system is a machine-based system designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments"
- The Act introduces the classification of AI systems based on risk to society. The Act outlines four risk levels:
- Unacceptable risk: An AI system that poses an unacceptable risk must be prohibited. Examples of these systems include AI systems driving facial recognition.
- High risk: AI systems that pose a high risk to society must comply with a range of requirements, including testing, data training, and cybersecurity, to ensure they comply with governing EU laws. Examples of these systems include automated insurance processing using AI.
- Transparency risk: These include limited-risk chatbots, deep fakes, and AI-generated content that must comply with transparency requirements.
- Minimal risk: These include common AI systems like spam filters and recommendation engines, which pose minimal risk to society and must follow currently applicable legislation, including GDPR.
 |
| Pyramid representing risk to society - 4 risk levels |
-
Fines for non-compliance: The severity of the infringement determines the fine. It can be up to 30 M euro or 6% of the total worldwide annual turnover, whichever is higher.
The EU AI Act lays the foundation for AI governance
Photo by Kindel Media from Pexels modified with Adobe Firefly
