Fenris Creations has appointed former Icelandic minister Áslaug Arna Sigurbjörnsdóttir as senior director for AI partnerships.
Sigurbjörnsdóttir will lead a new research initiative for Eve Frontier, collaborating with AI laboratories, universities, policymakers, and strategic partners.
Sigurbjörnsdóttir was elected to parliament in 2016 and became Iceland’s youngest Minister of Justice in 2019. She was appointed Minister of Higher Education, Science and Innovation in 2021, serving until December 2024.
She resigned as an MP in June after a leave of absence to attend Columbia University, which reinforced her belief that “public service experience can create value in many different arenas,” including the evolving field of artificial intelligence.
“When I left politics, I said I wanted to contribute to the changes AI is bringing,” Sigurbjörnsdóttir tells GamesIndustry.biz. “Having seen how policy is made and implemented, I know how badly decision-makers need evidence about how these systems actually behave, and this is where that evidence can be produced. I look forward to helping Fenris take the work forward responsibly.”
This research will examine how autonomous AI systems set objectives, adapt to changing conditions, and make decisions over extended periods.
“Autonomous AI agents are already moving out of research labs and into real economies, and we don’t know whether they stay aligned with what people actually want, or how they will behave once the stakes are real,” says Sigurbjörnsdóttir. “Agents are tested in simulations that end and reset, then deployed into a world that does neither. Fenris Creations has built one of the very few places on Earth where that gap can be closed.”
“Eve Frontier is designed around both human and autonomous participants, which immediately makes controlled research possible”
“Eve Frontier offers a set of conditions that are very hard to assemble artificially. It’s persistent and single-shard, one universe, one economy, one social graph, so prices are a genuine aggregate signal and reputation is global. How an agent behaves follows it everywhere, which is a precondition for studying trust at all. Information is imperfect at scale, task horizons are long, and outcomes are irreversible: mistakes stay made, debts stay owed, and every action consumes energy.”
Sigurbjörnsdóttir is particularly interested in how autonomous AI will respond to soft skills such as judgement, negotiation, and intent, noting that “those can be learned by playing repeated games with other humans and agents.”
“Nobody has ever observed a mixed human–agent society at scale because such a society hasn’t existed. The question is whether we study it somewhere it’s safe to get wrong, or wait and study it here.”
Sigurbjörnsdóttir says that, as Eve Frontier was designed with autonomous AI in mind, it’s the perfect testing ground for demonstrating how these agents behave.
“Unlike most virtual worlds, where automation is treated as a rule violation, Eve Frontier is designed around both human and autonomous participants, which immediately makes controlled research possible.”
The foundation of this research is Fenris’ partnership with Google DeepMind, announced as the developer transitioned from CCP Games following its $120 million sale to Pearl Abyss. Google also acquired a minority stake in the studio.
“I spent two days this week with teams from Fenris Creations and Google DeepMind, where we exchanged ideas and started working through where our practical collaboration goes next, and what stood out was how closely both sides already think alike about persistent worlds and agents,” Sigurbjörnsdóttir recalls.
“My remit is deliberately wider than any one partner. This has to be open to more than one lab, alongside universities and policy institutions”
“DeepMind’s own history is really the argument for this. Atari, AlphaGo, and StarCraft 2 established that games are a legitimate substrate for AI research, and they were right. But all of those environments reset. Eve doesn’t. That ladder runs out exactly where ours begins. My remit is deliberately wider than any one partner. This has to be open to more than one lab, alongside universities and policy institutions. If one company owns the proving ground, the evidence is worth less to everyone, including them.”
Sigurbjörnsdóttir says “some conversations have already started” with other partners, including frontier AI labs, academic researchers, AI safety institutes and regulators. “I hope that in the coming months we can open it up more widely.”
