Artificial intelligence: a systematic overview of the risk picture

In this note, we map risks associated with artificial intelligence (AI) and propose solutions. The aim is to form the basis for a more informed conversation about how Norway can help reduce risks and exploit the potential inherent in the technology.

The risks can be divided into three categories:

• Structural problems: KI can create systematic discrimination against vulnerable groups, economic transformation challenges, challenge democracy and trust, as well as cause destabilizing race dynamics between states and corporations.

• Misuse risk: KI can lower the threshold for malicious actors to carry out advanced cyber attacks or develop chemical or biological weapons. It could also strengthen the ability of states and corporations to monitor and manipulate individuals and democratic processes.

Accident risk: KI can increase the risk of serious accidents when used in critical infrastructure, autonomous weapons and nuclear weapons systems.

All the risks will increase with more powerful KI. We need governments, scientists and companies to put in an extraordinary effort to address a range of technical and regulatory issues before the models become more capable. This note points out a way forward for this skill.
‍