Finn Metz first joined the SCET community after participating in the Startup Semester at Berkeley program at UC Berkeley in the fall of 2024. Originally from Germany, he studied computer science at the Technical University of Munich, and studied at Copenhagen Business School before attending UC Berkeley.
Finn has dedicated himself to developing and investing in the next generation of security technologies in the age of AI. Today, he is the co-founder of Seldon Lab, an accelerator investing in hardware, cyber, and research startups in the realm of AI safety. The Seldon Lab team has raised $1.3 million in their first batch and has since sold to xAI and Anthropic.
We connected with Finn Metz to learn more about his journey, mission, and the future of AI safety.
The Journey to UC Berkeley
Prior to attending UC Berkeley, Finn had already gained meaningful experience investing in working with early-stage ventures. It wasn’t until he attended Copenhagen Business School that he discovered a deep passion for AI safety research.
After learning more from his peers and familiarizing himself with the research, Finn realized that the Bay Area, and more specifically UC Berkeley, represented a confluence of his research interests and professional goals. He proactively sought and received admission to the Startup Semester program, which is a student exchange program run by SCET that allows any student anywhere in the world to come to Berkeley for one semester to be immersed in entrepreneurship and SCET’s famous courses.”
“I really needed to be at the center of the world when it comes to the topics that I am passionate about.”
The Origins of Seldon Lab
Additionally, Finn held a position at Apart Research, a nonprofit lab focused on AI safety. It was here that he met his cofounder and conceived what would eventually become Seldon Lab. Though he found the research meaningful, he was surprised that the group was conducting its activities solely through a nonprofit lens.
Finn’s exposure to the venture ecosystem instilled in him a strong understanding of the startup lifecycle, from conception, scaling, and eventual exit. His unique background in both industry and academia led him to recognize a solution to address these issues.
“None of the solutions that they built really scaled beyond a couple of researchers in a lab. We need all of this new infrastructure and technology. We need a better way to build this.”
“For-profit organizations and startups seemed like the most powerful, highest-leverage tool that we had to push this forward. We need more for-profit startups working on these AI safety problems that we’ve already identified within the lab.”
Rather than founding his own venture, Finn ultimately decided that founding a startup accelerator had the potential to create the maximum impact.
“I had a choice: do I want to take one big shot at one startup, or do I want to push a lot of the folks I’ve been working together with in that direction and enable a lot more startups?”
Once the team received funding, Seldon Lab kicked off its first batch with four startups building AI safety tools. Since then, the team has backed Andon Labs, Asymmetric Security & Workshop Labs.
AI Safety: Why is it so important?
Seldon Labs is founded on the notion that “the world is unprepared for the arrival of superintelligence.” Finn predicts that, “as AI becomes the biggest industry, securing and controlling it will follow to become the second biggest.”
As Finn puts it, “the bottom line is, we’re bringing about this insanely powerful technology. The CEOs and the lab leaders agree that this is unprecedented—we’ve never built something like this before.”
While new AI capabilities create huge market opportunities, its rapid rise creates complex societal, geopolitical, and cyber and infrastructure-related challenges that must be addressed.
“Very few people care about all of the consequences that will come with bringing about this technology. There are so many new threats that will emerge.”
“We have all these different risks that we’re exposing ourselves to by ‘running fast and breaking things’ without thinking about the consequences.”
When it comes to tackling these challenges, Finn adopts a “techno-optimist” perspective.
“How can we solve those problems? What technologies and infrastructure can we build to address these new risks and threats?”
Finn highlights three issues he sees as most pressing today.
Finn highlights three issues he sees as most pressing today.
Oversight & Alignment Challenges
At Seldon Lab, Finn invests in startups that “expose critical blind spots and failure modes for AI labs.”
“As AI systems get more capable, we’re losing visibility into what they’re doing. We had Lucid Computing in batch 1 to build hardware-level GPU monitoring. They are building something like x-ray vision for AI compute. If you can’t see who’s running powerful models and what they’re running, you can’t govern them.”The team has also backed Andon Labs, a company that continues to make headlines for their experiments at xAI, Anthropic, and YC.
Shortcomings in Cyber Resilience
“When millions of AI agents hit the internet, the attack surface explodes. Our infrastructure wasn’t built for this. We need to secure an internet for humans, where AI attacks become the default.”
Finn firmly believes that building infrastructure for a safer internet is critical, and Seldon Lab has backed companies building in this space.“Companies we backed, like Asymmetric Security and 0labs are building defensive infrastructure for dramatic increases in cyber capabilities. Think cybersecurity 2.0, designed for AI-scale threats.”
Gradual Disempowerment
“As AI automates economic value and takes our jobs and agency, regular people lose bargaining power in society. You become more and more what you consume, and if you don’t provide economic value to the government/society, the government doesn’t need you anymore.”
Seldon Labs has backed Workshop Labs, a company building infrastructure to maintain the distribution of power as AI scales, including personal AI ownership models. Additionally, the team at Workshop Labs authored an article in TIME magazine describing these issues.
Building Responsibly in the Age of AI
Generally speaking, founders building AI products can adopt several practices to promote safe and ethical deployment of AI. Leaders can implement responsible scaling policies and to ensure that the AI cannot go off the rails.
Founders building AI safety native companies—the companies that Seldon Lab funds—must be concerned with the infrastructure itself. There are a plethora of questions to be asked:
“How do we make data centers robust? How do we deal with new cyber capabilities? What infrastructure do we need in order to deal with that? What kinds of new technologies do we need to handle deepfakes and misinformation on the web to make sure that our economy and communities work the same way?”
Reflections & Pearls of Wisdom
When asked what he loves most about his work, Finn noted the sheer scrappiness of a “zero to one” startup.
“Titles evaporate, and it’s just you solving a problem. And if you don’t solve it, your startup will die. All these systems of status that we’ve created don’t really matter. At an object level, it’s asking, ‘can you solve the problem from first principles?’ And I love that.”
To aspiring entrepreneurs, Finn highlights the importance of “swinging big” in times of uncertainty and change.
“A lot of things are changing rapidly. I think in these times of uncertainty, you should swing big. Make big, unprecedented bets on things. Try to build something entirely new or something that sounds crazy, and just see what happens. Nobody can predict the future.”
Moreover, he emphasized the importance of building for impact, rather than solely maximizing profit.
“Go for the problems that matter. Think about where you can have a positive influence, and where you can use your skills in a positive manner.”
What Lies Ahead
Finn acknowledged that, in a world defined by uncertainty, predicting the future is nearly impossible. He hopes to continue contributing to impactful startups and nurturing relationships with founders building at the frontier of AI. Eventually, he aspires to found a startup of his own.
Throughout his career, he hopes to continue meaningfully contributing to facilitating a smooth transition to a world with AI and mitigating pressing risks society faces today.
Finn Metz first joined the SCET community after participating in the Startup Semester at Berkeley program at UC Berkeley in the fall of 2024. Originally from Germany, he studied computer science at the Technical University of Munich, and studied at Copenhagen Business School before attending UC Berkeley.
Finn has dedicated himself to developing and investing in the next generation of security technologies in the age of AI. Today, he is the co-founder of Seldon Lab, an accelerator investing in hardware, cyber, and research startups in the realm of AI safety. The Seldon Lab team has raised $1.3 million in their first batch and has since sold to xAI and Anthropic.
We connected with Finn Metz to learn more about his journey, mission, and the future of AI safety.
The Journey to UC Berkeley
Prior to attending UC Berkeley, Finn had already gained meaningful experience investing in working with early-stage ventures. It wasn’t until he attended Copenhagen Business School that he discovered a deep passion for AI safety research.
After learning more from his peers and familiarizing himself with the research, Finn realized that the Bay Area, and more specifically UC Berkeley, represented a confluence of his research interests and professional goals. He proactively sought and received admission to the Startup Semester program, which is a student exchange program run by SCET that allows any student anywhere in the world to come to Berkeley for one semester to be immersed in entrepreneurship and SCET’s famous courses.”
“I really needed to be at the center of the world when it comes to the topics that I am passionate about.”
The Origins of Seldon Lab
Additionally, Finn held a position at Apart Research, a nonprofit lab focused on AI safety. It was here that he met his cofounder and conceived what would eventually become Seldon Lab. Though he found the research meaningful, he was surprised that the group was conducting its activities solely through a nonprofit lens.
Finn’s exposure to the venture ecosystem instilled in him a strong understanding of the startup lifecycle, from conception, scaling, and eventual exit. His unique background in both industry and academia led him to recognize a solution to address these issues.
“None of the solutions that they built really scaled beyond a couple of researchers in a lab. We need all of this new infrastructure and technology. We need a better way to build this.”
“For-profit organizations and startups seemed like the most powerful, highest-leverage tool that we had to push this forward. We need more for-profit startups working on these AI safety problems that we’ve already identified within the lab.”
Rather than founding his own venture, Finn ultimately decided that founding a startup accelerator had the potential to create the maximum impact.
“I had a choice: do I want to take one big shot at one startup, or do I want to push a lot of the folks I’ve been working together with in that direction and enable a lot more startups?”
Once the team received funding, Seldon Lab kicked off its first batch with four startups building AI safety tools. Since then, the team has backed Andon Labs, Asymmetric Security & Workshop Labs.
AI Safety: Why is it so important?
Seldon Labs is founded on the notion that “the world is unprepared for the arrival of superintelligence.” Finn predicts that, “as AI becomes the biggest industry, securing and controlling it will follow to become the second biggest.”
As Finn puts it, “the bottom line is, we’re bringing about this insanely powerful technology. The CEOs and the lab leaders agree that this is unprecedented—we’ve never built something like this before.”
While new AI capabilities create huge market opportunities, its rapid rise creates complex societal, geopolitical, and cyber and infrastructure-related challenges that must be addressed.
“Very few people care about all of the consequences that will come with bringing about this technology. There are so many new threats that will emerge.”
“We have all these different risks that we’re exposing ourselves to by ‘running fast and breaking things’ without thinking about the consequences.”
When it comes to tackling these challenges, Finn adopts a “techno-optimist” perspective.
“How can we solve those problems? What technologies and infrastructure can we build to address these new risks and threats?”
Finn highlights three issues he sees as most pressing today.
Finn highlights three issues he sees as most pressing today.
Oversight & Alignment Challenges
At Seldon Lab, Finn invests in startups that “expose critical blind spots and failure modes for AI labs.”
“As AI systems get more capable, we’re losing visibility into what they’re doing. We had Lucid Computing in batch 1 to build hardware-level GPU monitoring. They are building something like x-ray vision for AI compute. If you can’t see who’s running powerful models and what they’re running, you can’t govern them.”The team has also backed Andon Labs, a company that continues to make headlines for their experiments at xAI, Anthropic, and YC.
Shortcomings in Cyber Resilience
“When millions of AI agents hit the internet, the attack surface explodes. Our infrastructure wasn’t built for this. We need to secure an internet for humans, where AI attacks become the default.”
Finn firmly believes that building infrastructure for a safer internet is critical, and Seldon Lab has backed companies building in this space.“Companies we backed, like Asymmetric Security and 0labs are building defensive infrastructure for dramatic increases in cyber capabilities. Think cybersecurity 2.0, designed for AI-scale threats.”
Gradual Disempowerment
“As AI automates economic value and takes our jobs and agency, regular people lose bargaining power in society. You become more and more what you consume, and if you don’t provide economic value to the government/society, the government doesn’t need you anymore.”
Seldon Labs has backed Workshop Labs, a company building infrastructure to maintain the distribution of power as AI scales, including personal AI ownership models. Additionally, the team at Workshop Labs authored an article in TIME magazine describing these issues.
Building Responsibly in the Age of AI
Generally speaking, founders building AI products can adopt several practices to promote safe and ethical deployment of AI. Leaders can implement responsible scaling policies and to ensure that the AI cannot go off the rails.
Founders building AI safety native companies—the companies that Seldon Lab funds—must be concerned with the infrastructure itself. There are a plethora of questions to be asked:
“How do we make data centers robust? How do we deal with new cyber capabilities? What infrastructure do we need in order to deal with that? What kinds of new technologies do we need to handle deepfakes and misinformation on the web to make sure that our economy and communities work the same way?”
Reflections & Pearls of Wisdom
When asked what he loves most about his work, Finn noted the sheer scrappiness of a “zero to one” startup.
“Titles evaporate, and it’s just you solving a problem. And if you don’t solve it, your startup will die. All these systems of status that we’ve created don’t really matter. At an object level, it’s asking, ‘can you solve the problem from first principles?’ And I love that.”
To aspiring entrepreneurs, Finn highlights the importance of “swinging big” in times of uncertainty and change.
“A lot of things are changing rapidly. I think in these times of uncertainty, you should swing big. Make big, unprecedented bets on things. Try to build something entirely new or something that sounds crazy, and just see what happens. Nobody can predict the future.”
Moreover, he emphasized the importance of building for impact, rather than solely maximizing profit.
“Go for the problems that matter. Think about where you can have a positive influence, and where you can use your skills in a positive manner.”
What Lies Ahead
Finn acknowledged that, in a world defined by uncertainty, predicting the future is nearly impossible. He hopes to continue contributing to impactful startups and nurturing relationships with founders building at the frontier of AI. Eventually, he aspires to found a startup of his own.
Throughout his career, he hopes to continue meaningfully contributing to facilitating a smooth transition to a world with AI and mitigating pressing risks society faces today.