Transparent AI: What do we expect from it?

Discussions on artificial intelligence (AI) and its ever-increasing application to new areas intensify. While there is a lot of hype around it, new AI applications may be a real concern for the common people. If tasks that have been duties of humans are now taken care by intelligent autonomous systems, how can we be certain that such functions are completed diligently and safely. These concerns relate to our lack of understanding on how AI operates.

The EU’s Artificial Intelligence Act proposal (AIA) aims at unmasking AI. The objective is to generate human centric, secure, trustworthy and ethical AI. This article will analyze how in the context of AIA trust and accountability are created through rules on disclosure and transparency. Special focus will be on tension between rules on transparency and rules on the protection of trade secrets under the AIA. This conflict is also identified under the Trade Secrets Directive through its exceptions to trade secret protection. The article aims to identify situations where trade secret protection needs to yield due to these exceptions. The analysis will also cover rules that enable access to documents held by public authorities under the transparency regulations.