The first legal framework on AI – Is this for real?

No wonder, AI has made it possible to accomplish tasks that were once thought to be out of our league. Right from controlling the traffic to assisting the surgeons in performing the various medical procedures, AI has carved a niche for itself. What has always been a hot topic of discussion is the Regulation aspect of AI. On that note, The EU Commission adopted a proposal for the first legal framework on AI recently. This aims at imposing obligations on businesses across multiple sectors, including that of the life sciences (the Regulation).

AI, under the Regulation, is defined as the "software that is developed with one or more of [certain] approaches and techniques and can, for a given set of human-defined objectives, generate outputs such as content, predictions, recommendations, or decisions influencing the environments they interact with."

The legal framework that has recently made its presence felt has a crystal clear objective – to ensure a level of trust in AI systems that are made available. Well, that is not all. There is more to it. It adopts a risk-based approach.

What is the risk-based approach all about?

The risk-based approach throws light on the different risk categories. They are –

• Unacceptable risk – The term itself speaks a lot for itself. This is the risk that is a clear threat to individuals' safety, livelihood, or rights. A threat to all of this is highly unacceptable – hence the name. If any system is considered to cause a threat to any of the above, it would be banned, without a second thought.
• High risk – Evidently, the risk element is high. Some cases, as far as life sciences is concerned where this holds true is – administration of justice, types of safety components and products, remote biometric identification and categorization of people, etc. All those AI systems that fall under this category have a set of obligations to comply. Some of them are –

1. Adequate and clear information to be provided to the user.
2. Adopt the required oversight measures to minimize the errors.
3. Check on the robustness, security as well as accuracy.
4. Trace the results by logging the activities.
5. Thorough and detailed documentation.
6. Conduct sufficient risk assessments to deal with the undesired circumstances in a better manner.

All of these apply to the providers of high-risk AI systems as well as the manufacturers of products that include high-risk AI systems. As far as the distributors, importers, users and other third parties are concerned, will be subject to the provider's obligations if –

1. They trademark or modify an existing system
2. They place a high-risk AI system on the market
3. They place a high-risk AI system into service under their name

• Limited risk – This is where the AI systems are well aware of the transparency obligations that they have to comply to.
• Minimal risk – This category is where most of the AI applications fall under. The reason why it is named as minimal risk is that the applications pose minimal to the low risk to individuals' rights or safety.

In a nutshell, the Regulation is all set to have a significant impact across all sectors. Since the complete legislative process is not yet done, the businesses will now have to wait to see how the principles set out in the Regulation.