The CEO of Swedish vibe coding startup Lovable has promised to spend 50% of his future earnings combating ‘bad AI’.
Concerns over AI safety have been a focal point of policy debates across the US, the EU and the UK, with some raising concerns over potentially existential threats, as well as large-scale impacts of social profiling and the use of the technology by law enforcement agencies and militaries.
Speaking on stage at Nordic Tech Week in central Stockholm, cofounder Anton Osika said he would commit a large portion of his future earnings to improving AI safety, via Founders Pledge.
“If I'm going to work extremely hard for a company, rather than solving a major societal problem — which I also considered — then I have to feel that it will leave a positive mark on the world.”
He added, “Our mission is about empowering more people to have a chance to be part of the tech industry. Another aspect is that my cofounder and I have a founder pledge to give away 50% of our proceeds from what we make as a company to a philanthropic cause.”
In July, Lovable raised $200m in a Series A led by Accel to expand its engineering team and accelerate global growth. Last month, it reportedly received funding offers valuing it at $4bn — but Osika says the company’s ambitions go far beyond that.
“We want to be the first trillion-dollar company from Europe. It just means you have to have the right people working hard on the right things. And then we will achieve that.”
Often described as the fastest-growing company in the world, Lovable hit $130m in annual recurring revenue (ARR) last month. It now employs around 70 people, with plans to hire another 30 by the end of the year.
But the transition to more advanced AI is something that Osika cares about.
“AI is going to change human history forever — in some ways for the better, in some ways for the worse,” he says.
“For me, [the pledge] is specifically focused on how humanity handles the transition to artificial intelligence, and there are a lot of risks.”
He warned of scenarios where AI could misinterpret instructions with devastating consequences.
“At some point in the future, you could imagine giving some kind of instructions to an AI system, and then it goes out and executes them in a way that you didn't expect.
“Let's say I’m in charge of our Swedish Armed Forces AI system and I ask it to make sure that no Swedish people are killed by foreign forces. If the AI thinks that humans are extremely slow, the most reliable way to ensure that Swedish fleets are not attacked by foreign forces is, theoretically, to take out all the foreign forces. But that's not what you intended as a human.
“As AI becomes more and more powerful, it also becomes more unpredictable in what it will do when humans, who are not allies, both have access to extremely powerful systems. So I think that's the core of what I see as the problem.”
