What do you know about the AI Act?

AI has become indispensable in our society. Nearly every sector actively employs AI systems and tools, which are often highly beneficial. However, they can also pose risks, such as infringing on the privacy of citizens. The AI Act was therefore introduced to mitigate these negative consequences. Read on to learn more. 

Where doest the AI Act apply?

The AI Act establishes specific regulations for AI applications, including systems used for recruitment and personnel selection, as well as AI involved in decisions regarding promotions. Beyond human interaction, AI may also be employed in managing traffic safety and water resources, where the risk of negative consequences from improper application is significant.

Negotiations on the AI Act were concluded late last year after intense discussions, particularly concerning the list of prohibited practices and permitted applications. Member states and the European Parliament held differing views on these matters. European AI companies actively lobbied and cautioned that overly stringent regulations could have adverse effects.

Privacy violation

The EU's AI Act prohibits specific AI applications that may violate the rights of EU residents, such as:

• Uncontrolled collection of facial photos from the internet for facial recognition,
• AI that detects emotions in the workplace or at school,
• Manipulation of human behavior,
• Social scoring programs similar to those used in China, and categorization based on sexual orientation or religion.

The law also imposes new rules on law enforcement regarding the use of biometric identification systems (RBI). Real-time RBI may only be used in urgent situations, such as locating a missing person or preventing an attack. Permission from a judge is required for the use of this technology in other circumstances.

Transparency within AI systems

AI systems such as Chat-GPT must also become more transparent under EU regulations. It should be clear when content is generated by AI. Systems posing a high risk, where potential harm from errors could be significant, require additional measures such as continuous human oversight. Examples include education and healthcare systems.

The law is expected to be officially published in May 2024, after which the rules will gradually come into effect. Prohibitions will take effect six months later. In May 2025, the rules will apply to tech companies, with full implementation expected by 2026.

Want to get in touch? Fill out the form below.