A new regulatory framework for AI use

Within the regulatory framework, the risks of using AI are classified into four categories, namely unacceptable risk, high-risk, limited risk, and minimal risk. For “high risk” applications of AI, it will be mandatory that they meet certain requirements before they are able to be used on the market. For example, there will have to be adequate systems in place to assess and mitigate the risks in question, the data being fed into these systems will need to be of high quality to reduce the likelihood of discriminatory results, and appropriate human oversight will need to be ensured. Moreover, clear and accurate information must be provided to the user, and these AI systems must have a high degree of robustness, security, and accuracy. In addition to the regulation, the proposal also outlines the creation of a European Artificial Intelligence Board, which will be responsible for facilitating the development of standards for AI.

Under the AI activities deemed ‘high risk’, the Commission lists the evaluation of creditworthiness, as well as AI use in critical infrastructure like transport. The inclusion of these AI applications in the ‘high risk’ category has the potential to impact our industry, with the most immediate impact likely in the context of AI and creditworthiness assessments, but the long-term impact on the transport industry being equally important. It will ultimately require operational changes in assessing customers’ creditworthiness in instances where this is currently done through AI, possibly increasing waiting times for customers to be approved and requiring additional compliance measures to be accommodated.

Last month, the Members of the AIDA Special Committee and the TRAN Committee at the European Parliament held a joint hearing on AI and transport. Representatives from academia, the private sector and civil society shared their views on the long-term impact of AI on transport modes and mobility. A key theme in these discussions was the implications of self-driving vehicles, with most participants noting that the uptake of autonomous vehicles would likely negatively impact both safety and road congestion in the short term, whilst ultimately improving both in the medium-long term.

The Council of the EU also debated the proposal recently, with telecommunications ministers holding their first in-depth policy debate on this. Ministers welcomed the risk-based approach of the proposal, whilst noting that some key provisions require further clarification (such as a number of key terms not being well defined, and issues surrounding the scope of the act). The Ministers also spoke on the usefulness of regulatory sandboxes and experimentation to help ensure the AI legislative framework is fit for purpose, and the importance of effective enforcement and supervision, especially as it pertains to human oversight requirements. Once the European Parliament and Council are happy with the content of the new framework, it will then be entered into EU law.