🦾 Free AI for all Swedes, proposes the AI Commission

My view on the AI Commission's roadmap

This year, I've had the great privilege of being a member of the Swedish government's AI Commission. Now the commission has completed its work, and here's my personal view on the results and a small insight into my and the commission's work.

Some of our proposals are of course specific to Sweden, but the same logic can be applied to other countries and also companies, cities and even yourself.

Besides doing a good job overall, I had three priorities when I joined the commission. Two succeeded, one failed. I'll summarize this at the end of this text.

Time for Sweden to enter crisis mode

The AI Commission has now submitted its roadmap to Minister Erik Slottner. We did this seven months early. Originally, we had until summer 2025, but we quickly realized we were in much more of a hurry than that.

The situation is serious. Sweden missed the start of the AI era. Other countries were much quicker out of the starting blocks. But not only that, we're falling further behind. In a major global AI index, Sweden dropped from 17th to 25th place, and AI usage among Swedes didn't increase from 2023 to 2024. Sweden usually belongs to the top nations, so this is particularly serious.

During the internet revolution, Sweden was actually the best in the world for a while. So now it's time to pick up speed. Hopefully, the AI Commission's report will be the starting gun for Sweden in the AI race.

One of the AI Commission's proposals is for Sweden's government to enter something resembling crisis mode:

"We therefore propose that a special task force be established at the Prime Minister's Office in the Government Offices, to monitor that necessary AI measures are implemented."

AI for all reform - free AI for every Swede

However, the race isn't over. It has just begun, and Sweden has the chance to take the lead, even if we were slow at the start.

Sweden has an advantage. We're good at broad transitions where the whole society comes along – and that happens to be the key to success in the AI era. Here's what the report's introduction states:

"Now we can control machines with our natural language and get help creating new material in the form of text, images, program code, sound, and much more. Here, our human intelligence is combined with artificial intelligence. This means we can work faster and with higher quality while becoming more creative and innovative. AI has thus become a tool for us all."

AI unleashes human potential and creativity. The country where most people use AI will thus unleash a massive wave of innovation. Sweden should therefore become the country in the world where most people use AI, just as we were once the country where most people had home PCs and fast internet.

We'll get there through an AI for all reform. The AI Commission proposes that all Swedes get free access to AI tools like ChatGPT, Gemini, Claude, and many others. This will be done through an AI hub, where Swedes can log in and get access to paid versions of AI tools for a limited period.

But don't most AI tools have good free versions? That's true, but you don't get access to all features and quickly hit the usage limits. We want Swedes to use AI all the time. Additionally, we don't know how generous AI companies will be in the coming years and how long they'll continue to offer free alternatives. Most importantly, it's a very strong signal to all Swedes that this is something important - and for everyone. AI is no longer just for engineers and computer scientists, but for all of us.

If we could, through magic, get everyone in Sweden to use AI for the right tasks, we would see an unprecedented increase in productivity. We would work faster and with higher quality. Swedish companies would have a strong competitive advantage and the public sector could offer citizens better service and save billions. That's the expected outcome. Magic doesn't exist, but an AI for all reform is the closest thing to a magic spell we can get.

Then we have the unexpected too. No one could predict that one effect of the subsidized home PCs and Sweden's rapid broadband expansion would be Spotify. But when Spotify founder Daniel Ek surfed the fast internet, before most people in the world, instead of the incredibly slow modem internet, his mind opened to the possibilities of streaming. Such things can't be predicted in advance.

What it will be this time, we don't know. New forms of expression, billion-dollar companies, ways of communicating, art forms, research breakthroughs, better urban planning, and new forms of democracy?

If Sweden becomes the country in the world where the use of generative AI increases fastest then a high proportion of all these new things will be created here.

A little bit of school

According to the government directives, the AI Commission was not allowed to make proposals for schools. Many have reacted against this, myself included. But writing about schools wasn't forbidden, so we did that.

Among other things, we highlight the possibility of every student getting an AI private tutor. Like what Khan Academy is doing in the US with Khanmigo. A teacher that's there 24/7 and helps the student step by step. An excellent complement and help to the human teacher.

No doomsday prophecies

Another part that I'm particularly satisfied with concerns the section on existential risks with AI.

"The AI Commission believes that concern about the so-called existential risk should be considered, but the commission also regrets that parts of the public discussion about safety have tended to focus on speculative risks with very low probability. As we touch upon in the previous section, it's important that society's work to handle AI-related security risks is evidence-based."

Of course there are risks with AI, which are addressed in the report, but the existential risk specifically is bad science fiction. Now it exists in the general debate about AI, which creates concern that must be taken seriously. But the commission thinks the focus should be on the real threats, not "on speculative risks with very low probability."

Working in the AI Commission

Despite the shortened time, we've accomplished quite a lot. The commission itself has had five full-day meetings and about 30 digital meetings. Beyond that, we've had over 200 meetings with more than 150 actors from large parts of Swedish society. (I haven't participated in all these meetings, a large and impressive load has been carried by the chairman together with the secretariat.)

At our first meeting, chairman Carl-Henric Svanberg said that all members would both have the opportunity and be expected to actively contribute to the report. All chairmen say that, and sometimes it's true. Sometimes they want to dictate everything themselves, despite the inclusive words. So I waited to see if it would really be as he said. But that's exactly how it's been. Carl-Henric has focused on getting us to perform as a group, created a good atmosphere, and in that way led us forward. Together with the secretariat, he has been very active in meeting all societal actors. He was also a driving force in the important decision to advance the report.

Based on interest and expert knowledge, we members have had extra responsibility for different parts of the report. Together with Fredrik Heintz, Ulrika Lindskog, and Sylvia Schwaag Serger, I've worked extra on the AI for all chapter, which includes public education and the free AI reform. We've written texts that were then processed by the secretariat. Of course, we've always had the opportunity to have input on other areas.

What I succeeded and failed at

Since I've spent a lot of time studying Sweden's IT miracle in the 90s, I saw that a similar recipe could work now too. Then, as now, the key was increased usage. My first priority was to find a well-functioning proposal that gave Swedes free AI. Was there a home PC reform for the 2020s? I had thoughts about an AI deduction or an AI check. Together with colleagues working on AI for all, we concluded that the best solution is a hub where you can easily log in and access AI tools.

My second priority was the risk section. I wanted it to focus on real risks, not the bad science fiction risks that have dominated the debate. This wasn't a problem at all. We had several people knowledgeable about real risks, and I think we achieved an overall very good risk section.

Finally, my failure. I had hoped to change the government's directive so the commission could make proposals for schools. As mentioned, that didn't happen. Even though we wrote some about schools, our roadmap would have been better if we could have met with school representatives and submitted a handful of strong proposals to the government.

Separation anxiety

The commission has been incredibly rewarding and I feel clear separation anxiety. The combined knowledge in this group is impressive, I've learned so much. There have been many emails. Every Thursday at lunch, we've connected and discussed the report. Now there's an empty gap in the calendar.

Now begins the equally important work of getting all parts of Swedish society to act. Preferably implement what we propose, but mainly just get started. The opportunities are enormous. We can give Sweden an advantage that lasts for generations. There's nothing to wait for.

Mathias Sundin
Member of the AI Commission

There's so much to write about the report. It's available to read on the government's website, but I've also made it into a GPT, so you can chat with the report about the report. Note, it's not an official product from the commission and the same strengths and weaknesses as with generative AI in general apply to this GPT as well.

What I've previously written about the AI Commission:

🦾 On the AI Commission: Your country should become a centaur society

🦾 What can we learn from the Swedish IT miracle for the AI era?

🦾 Boosting AI usage - report from the first meeting of the AI Commission

🖥️ Should Sweden invest in models or increase the use of AI? – report from the second meeting of the AI Commission