The world’s best security should belong to everyone.

We believe everyone should have access to the best security, as easily as possible. We’re setting a new standard for transparency in the security industry by making our harness open source and our research open to scrutiny. The security we develop for demanding work in defence and critical infrastructure should be available to the developer shipping alone, the startup finding its feet and the enterprise running systems people depend on.

We’re already advancing novel vulnerability research. We’ve created the first self-evolving, multi-model cybersecurity harness and made it open source. With it, we’re finding previously unknown flaws in software running on billions of devices, smartphones and servers around the world (such as the Linux kernel). Our public disclosures and upstream fixes let anyone check that work. Now we’re turning that research into security more people can use.

Software already writes itself. It should secure itself, too. AI is changing how we build software. We believe it will also change how we interact with devices and operating systems. We see a future where software is generated on demand, adapting as we use it, and software factories build, test and improve it continuously.

We don’t want to give people another security tool to operate. We believe that model will become obsolete as software takes on more of its own development. Security should be part of how software works and evolves, finding and fixing vulnerabilities as it changes. That’s the future we’re building for: software that secures itself, so people can focus on what they want to create.

We’re backed by Y Combinator. Its idea of making something people want fits what drives us: give people the freedom to build, with security they can actually use. We began in Zurich . San Francisco pushed us to think bigger . Now we’re building for the world .

Research first.
Built for the real world.

The Swiss Applied AI & Cybersecurity Research Lab
Explore our research philosophyClose our research philosophy

Adding “AI” to a product page is easy. Advancing what it can actually do takes research. We’re at the forefront of applying AI to novel vulnerability research, finding previously unknown weaknesses and developing better ways to verify and fix them. We turn what we learn into tools people can use. Real problems from industry shape the work. Self-securing software remains the goal.

We collaborate with leading AI labs and hyperscalers, including Microsoft, as well as defence organizations. Our work with industry leaders, including EY (Ernst & Young), keeps the research grounded in the problems companies face today.

We work across frontier and open-weight models. We don’t believe any single model is enough to find everything. Different models bring different strengths. Our approach is multi-model: test models against real security problems, combine what works, and keep improving.

Our roots are in Zurich, where we began and where our offices sit right next to ETH. We work closely with leading AI labs in San Francisco and collaborators around the world. Swiss roots, independent judgment, global ambition. We choose models for their capabilities, not allegiance to a vendor, country or geopolitical agenda.

That vision depends on open research: reproducible results and tools people can inspect, question and build on. Sharing the work lets others improve it and put it to use in ways we haven’t imagined.

Security you don’t have to think about.

People want to build. Security too often gives them another dashboard, another report and more work.

We want to build security people love using because it quietly takes care of the work. It lives inside the development pipeline, finding and fixing as software changes.

For developers, engineers, researchers, enterprise teams and fast moving tech companies. Whoever you are, whatever you build, security should give you time back.

Software already builds itself.
Now it can secure itself, too.

That’s the future we’re building for. AI changed the pace of building. Security needs the same leap.

Another report won’t get us there.

For too long, security has arrived as a separate engagement. A point-in-time test. A list of findings. Work handed back to the people trying to build.

We’re bringing security inside the way software gets made.

Find. Fix. Verify. Keep building.

Our vision is a continuous loop: find vulnerabilities, establish what is exploitable, auto-fix the problem, and verify the result. Part of development from the start, evolving with every change.

Your workflow is the interface.

The best security tool is one you don’t have to think about. It works where you and your agents already build. No more dashboards to babysit. No constant context switching.

You stay in control. Your team sets the scope, permissions and conditions for putting changes to work.

Agents need security tools, too.

People are already building software by directing teams of agents. That shift is accelerating. Those agents need specialized systems: tools, context and verification built for the job.

Our conviction is that better models make that foundation more valuable. We’re building the cybersecurity harness for that future, for agents to use directly and people to direct.

Security for software factories.

Today, people direct the agents. Tomorrow, more of the building, operating and improving will happen within software itself. Security has to be part of that loop. We’re creating the technology that makes it possible for software to secure itself.

Open it. Question it. Build on it.

Builders should be able to inspect the tools they depend on. Our security harness is open source. Read it, run it, extend it. Bring a better idea.

We share what we learn through public research and documentation. Disclosures are coordinated with maintainers. Client work stays private.

For those who push the world forward.

From your first idea to a startup’s next release, a bank’s core systems, or the infrastructure the world depends on. Security should give you the confidence and freedom to build what comes next.

Talk to us

Team