What is the future of Artificial Intelligence regulation?

Source Domain: www.telefonica.com

Reflection on the future of Artificial Intelligence regulation is conditioned by the fact that this technology is advancing at an extraordinarily rapid pace. Today we know that we are moving towards AI that is increasingly closer to human intelligence, agentive, contextual, emotional and culturally shaped.

An AI that will act as an invisible, omnipresent infrastructure, integrated into our daily routines, and that will function as a cognitive companion capable of making decisions on our behalf. In this scenario, regulation should not be based on how it works, regulating according to its internal processes, but rather on the consequences it may produce, and therefore it must become much more dynamic, technical and continuous.

Regulation will have to adapt to a reality where AI is not a product, but an infrastructure. Supervision must be permanent, based on real-time data and automated auditing using algorithms capable of monitoring and explaining other algorithms. The requirements of transparency, traceability, natural explainability and continuous risk assessment must form the basis of the new regulatory framework.

It is important to raise the level of discourse on risks, looking not only at the micro level, but also at the macro level: society, culture, politics, democracy and also the individual as a free agent. At the same time, equal and non-discriminatory access to technology must be provided if we do not want to have first-, second- or third-class citizens in areas such as agentive AI or neurotechnology.

Differences in regulation between different countries or regions

Differences between regions reflect different views on the role of the state, technology and fundamental rights. The European Union promotes a more protective framework, focused on the protection of individuals and risk management; the United States maintains a sectoral approach, more dependent on private innovation; China focuses on a strongly centralised model, oriented towards control,…

Source