Now that the EU legislators have arrived at a political agreement on several controversial aspects of the EU AI Act, Ilya Billig from the Gen AI Reality Check channel and me sat down to discuss the act and its implications. Here is an excerpt of the interview.
Ilya: Welcome back, Biju. Let's start with the EU AI Act. What does this act cover?
Biju: The EU AI Act covers AI systems in general, aiming to protect our fundamental rights from the risks posed by autonomous AI systems making decisions on our behalf.
Ilya: Is there a classification of different AI systems in the Act?
Biju: Yes, the Act classifies systems into three categories: low or no risk, high risk, and prohibited systems. Examples include spam filtering for low risk and biometric scanning for prohibited systems, with some exceptions for law enforcement.
Ilya: Are there penalties for violating this Act?
Biju: Violations can lead to fines up to 35 million euros or 7% of the total worldwide annual turnover for the preceding financial year.
Ilya: Does the Act apply only to European companies?
Biju: It applies to every company that has business in Europe, similar to the GDPR's extraterritorial reach.
Ilya: When will the EU AI Act be enacted?
Biju: The political agreement has been reached, and we expect the law to be ratified by June next year, with companies having two more years to prepare.
Ilya: What should companies start doing to prepare?
Biju: Companies should assess their level of AI governance and understand how AI ethics are incorporated into their applications. They should also classify their systems according to the Act's categories.
Ilya: Are there companies or people that can help with this preparation?
Biju: Yes, there are guidelines for conformity assessments and organizations like IEEE CertifAIEd and TÜV that offer services to help companies prepare.
Ilya: Will the EU AI Act put Europe at a technological disadvantage?
Biju: In the short term, there might be a disadvantage as investors may shy away. However, in the long term, products from the EU could be seen as safer and of higher quality.
Ilya: What incentives are being discussed alongside the Act?
Biju: The EU is proposing incentives like a Sandbox for startups to test high-risk AI systems and financial backing for startups to promote innovation.
Ilya: Who will benefit from this Act?
Biju: The Act aims to benefit society by protecting fundamental rights. Additionally, assessors, legislators, certification bodies, and companies providing related services will benefit.
Ilya: Thank you, Biju, for discussing this important topic of the EU AI Act.
Watch the full interview on YouTube or click the video below.