The European Union’s AI Act is a landmark legislation that marks the first major attempt to regulate Artificial Intelligence (AI) on a global scale. With its far-reaching implications, it’s criticality for businesses, developers, and stakeholders, in this article, we focus on key things that you need to know as someone who is interested in understanding the EU AI Act.
EU AI Act: Five Things You Need To Know
1. Risk-Based Approach
The EU AI Act is a risk-based regulatory framework. Artificial Intelligence (AI) systems are classified into four risk categories:
- unacceptable risk,
- high risk,
- limited risk, and
- minimal risk.
This classification dictates the level of regulatory scrutiny and compliance requirements that an AI system will be subjected to. Of course, the high-risk applications face the most stringent controls. This approach allows for flexibility and innovation (especially in lower-risk AI applications) while ensuring that those posing significant risks to safety and fundamental rights are adequately regulated.
Make no mistake, that the development and deployment of AI systems that can pose unacceptable risk shall be prohibited as per this law. These include AI systems that deploy subliminal techniques to manipulate persons to their detriment, exploit vulnerabilities of specific groups, conduct social scoring, or use real-time biometric identification in public spaces for law enforcement, with certain exceptions.
Tip: Evaluate the risk posed by your AI systems to ensure that your products comply with the EU’s standards.
2. Obligations for High Risk Systems
While systems that have the potential to violate rights or pose significant risks to individuals’ safety are considered to pose unacceptable risk and will be prohibited, the systems that will create high risk will be subject to all requirements from this regulation. These requirements include data governance, transparency, and explain ability i.e., provision of clear information to users; robustness and accuracy; human oversight; and specific documentation requirements such as logging, adopting extensive testing methodologies and record-keeping to enable traceability.
Tip: Businesses operating or aiming to launch AI solutions that may pose high risk must prepare to meet all these obligations.
3. Transparency Requirements
For AI applications that may not be classified as high-risk, particularly those interacting directly with consumers (like chatbots), must adhere to transparency requirements. This includes providing users with information that an AI system is being used, ensuring that users are aware of their interaction with an AI and not a human. This transparency is vital for maintaining trust and integrity in AI systems across all sectors.
Tip: Businesses are recommended to implement transparency requirements into all AI systems irrespective of the risk.
4. No Requirements for low risk applications
While many will talk about the doom and gloom that this regulation will bring, it is essential to share that AI systems that are deemed to pose a low or limited risk shall have no requirements being imposed by this legislation..
Tip: We recommend to formally evidence whether your AI system poses a minimal or no risk and not assume this.
5. Enforcement and Penalties
Like the EU GDPR, the EU AI Act will also be enforced rigorously. There are significant penalties for non-compliance. Depending on the violation, fines can go up to €30 million or 6% of the total worldwide annual turnover of the preceding financial year, whichever is higher. This strict enforcement underlines the EU’s commitment to ethical AI development and usage. It also signals the importance of compliance for all AI stakeholders.
Tip: We recommend you as a company to start now. And, we can help you.
Conclusion
In my opinion, the EU AI Act is a landmark regulation that will shape the future of AI development and usage not only within the European Union but also globally, as businesses and organizations worldwide will need to comply with these regulations to operate in the EU market. I believe many countries will now go for similar regulations. So, as business you need to start strengthening your risk management, data governance, transparency, testing and logging capabilities to name a few.
While it creates work for companies, I am convinced that this will contribute to a safer, more ethical AI landscape that is good for us humans.
