Ten measures for the new AI minister

Congratulations to Karianne Tung, our newly appointed Minister of Digitalization. With the appointment, the Government shows that it understands the need for an overall management of the rapid development of artificial intelligence (AI), as Inga Strümke and I have in demand in Aftenposten.

There are plenty of problems and challenges that Tung has to deal with. We at the Centre for Long-Term Policy therefore propose ten concrete measures that not only reduce the risks of using AI, but also give us the opportunity to take advantage of the technology.

• Parliamentary message
  Norway must as soon as possible develop a transparent and predictable regulatory framework that clarifies responsibilities and facilitates the safe and ethical use of IT in the public and private sectors. It requires a general announcement in which the government sets out its overall AI strategy, where they make clear where they want to go with the technology and how they will facilitate the use of AI for society. We can't just wait for the EU.
• Turn Digdir into a KI hub
  The ambition of the Directorate for Digitisation to be a world leader in digitisation should be extended to include the use of artificial intelligence. The Directorate should act as a competence bank for artificial intelligence in the public sector, and its aim must be to raise the quality and usability of public services.
• Empowering workers
  KI can be both a gift and a threat to workers. Authorities should follow developments closely and create a plan to ensure that KI strengthens workers' rights and well-being in different scenarios.
• Ki-register
  Sensible regulation requires transparency and transparency. In any case, a register for the use of AI systems in sensitive areas such as health, critical infrastructure and finance will become a requirement of the EU and should be implemented as soon as possible. We should also introduce a register for accidents that may be caused by KI. In the event of serious accidents, a body à la Accident Investigation Board should investigate the accident.
• The KI registers will not put a lid on innovation, but rather ensure sufficient confidence that KI is used in a prudent manner.
• Algorithm supervision
  However, registration is not enough. We need an algorithm oversight to ensure that KI systems do not violate Norwegian law and people's fundamental political, civil and economic rights. Supervision will give confidence and can ensure that algorithms contribute to more accurate, fair and transparent decisions.
• Such a body should form part of an existing competence environment. The inspectorate, given that they are given a broader mandate than they have today, could be such an environment.
• Data sharing
  In order for AI systems to be as good as we want, they need to have access to a large amount of data. To prevent systems from giving unfair results to some groups, they must be trained on data from and about the entire population, especially minority groups. However, it requires more data to be collected and shared.
• Authorities have already made efforts to facilitate data sharing, but more needs to be done. There are still a number of restrictions on the use of health data that need to become less square. We need to manage data more like a common resource and require companies that collect data about people to have to share the information with others. And if we are to train large language models on Norwegian data, we need a new policy to recognise people's intellectual property and compensate data producers for their work without stopping all development.
• Know your customers
  Transparency, sharing and accountability clarification are necessary factors to ensure prudent and trust-building implementation of KI. But that's not sufficient. KI can bring big gains, but it also comes with risk challenges.
• Powerful AI systems can strengthen the capacity of malicious actors and thus make us more vulnerable to terrorist attacks. We should consider the possibility of requiring companies that offer the most powerful KI systems to know their customers, similar KYC requirements in finance.
• Digital Certification Systems
  Deepfakes can be used for revenge porn and for libelous purposes, but can also undermine the public conversation. If we can no longer be trusted to know who is saying what in public, the trust democracy stands on is undermined. We need more Ki-robust certification systems that help us distinguish between fake and genuine content. >
• Security-NOU
  Besides the risks associated with malicious actors and deepfakes, there is also a danger that AI systems could contribute to catastrophic accidents. One danger is that KI to be integrated into nuclear weapons systems. Another is that the systems become so powerful and autonomous that they can constitute a existential threat to humanity.
• We need a public inquiry to get an overview of this threat picture and what measures we can take to mitigate the risk.
• International participation
  Much of the work to ensure safe KI has to take place internationally. The technology comes from large companies in other countries and there is a danger that a race between the United States and China or between the top AI companies will lead to security being sacrificed to win the race.
• Such a race can only be stopped with international cooperation and Norway must take part in this cooperation. Tung should therefore work with the Ministry of Foreign Affairs to establish a separate department at KI in the section for global security.
  
  The government should also consider how Norway can prevent powerful KI from being used for illegitimate warfare and terror. One possibility is to impose export controls on powerful KI systems that can be used for such purposes.

It is time to set the premises for how technology will affect society. We hope our new Minister for Digitalization seizes the opportunity.

‍