Horizon3 Raises $250M to Prove What Attackers Can Exploit

A proactive security startup led by the ex-CIO of GE Capital raised $250 million to expand its ability to test both on-premises and cloud-based infrastructure.

See Also: Why an AI Harness May Matter More Than the Latest Model

The NightDragon and NEA-led Series E funding will help San Francisco-based Horizon3 test how an attacker could move across infrastructure, identity and web applications, said Chief Revenue Officer Matthew Hartley. He said Horizon3’s strategy is to demonstrate what can actually be exploited rather than simply identify potential vulnerabilities.

“We’re oversubscribed on the round, which is indicative that people around the investment community wanted to participate in something that I would call applied artificial intelligence that’s actually working,” Hartley told ISMG. “The feedback during the fundraise was, ‘Hey, you guys are solving a real problem. You’re using AI in a safe way that adds a lot of value to your customers and we want to be a part of that.'”

Horizon3, founded in 2019, employs 506 people and has raised nearly $430 million. The firm notched a $2 billion valuation in conjunction with the Series E funding, tripling its valuation from $650 million at Series D in May 2025. The company has been led since its founding by Snehal Antani, who spent nearly four years as CTO of JSOC, nearly two years as CTO at Splunk and nearly three years as CIO of GE Capital (see: The Three Hardest Parts of Being a Security Leader).

How Defenders Benefit From an Attacker’s Perspective

Hartley said Horizon3 gives defenders an attacker’s perspective to demonstrate which vulnerabilities can actually be exploited and what an attacker could accomplish afterward. Horizon3 is working with Anthropic to study how attackers can use AI against infrastructure and how Horizon3 can incorporate comparable capabilities into its own testing.

“We address the one problem everybody’s got, which is infrastructure attack,” Hartley said. “Regardless of how you initially access an environment, you’re going to work and pivot and move laterally through infrastructure. Horizon3 is the best in the world, bar none, at infrastructure and we can keep going deeper and deeper into that, both on-prem and cloud-based infrastructure.”

The company has invested more than $100 million in production safety through laboratories, test ranges and other testing infrastructure, and Hartley said some of the Series E proceeds will expand those tools. Horizon3 must demonstrate exploitation without interrupting the customer’s business, and achieving that requires regression testing and a catalog of hardware, operating systems and software revisions.

“We’ve spent over $100 million in production safety through the construction of labs, test ranges and various types of scaled-up capability, and we’re going to be using portions of the raise to scale this even further, to test every piece of infrastructure that exists, copies of that, older versions of operating systems,” Hartley said.

Web application capabilities had been among customers’ most frequently requested features, Hartley said. Combining that with infrastructure and identity testing will help organizations examine how an attacker obtains initial access through an application, what the attacker can reach after gaining access, how identity can be exploited and whether the attacker can ultimately reach critical systems or data.

“Customers no longer can run manual methods in the past,” Hartley said. “They need something that’s automated, and they need an attacker’s perspective. Just because you have all these vulnerabilities, which are those that could be exploited? And that’s where Horizon3 comes in. We can prove that exploit at a scale nobody else can.”

How Horizon3 Goes Beyond Automating Vulnerability Scanning

Attackers can use LLMs to analyze existing libraries, repositories and other datasets at a speed and scale that were previously impractical. Older web applications that have not been well-maintained or updated can therefore become increasingly attractive targets. Organizations therefore face greater risks from existing applications because adversaries have more efficient tools for finding weaknesses in them, he said.

“Bad guys got a lot of weapons at their disposal to go train on that data, put it to work on legacy web apps, and again, they haven’t been written as well. They haven’t been updated. Now they can attack at a scale and speed they couldn’t before,” Hartley said.

He recommends starting with external assessment to identify exposed applications and infrastructure, then determine whether exploitable flaws can provide initial access. From there, organizations need to examine lateral movement, identity exploitation and paths to their crown jewels. AI compounds the issue because adversaries can simultaneously test numerous potential entry points at machine speed.

“The problem for our defenders, our customers, is that with AI, you can test all of that at machine speed,” Hartley said. “Now you don’t need a human being anymore just on the web app or just looking for an unpatched firewall. The days of human in the loop are over. You’ve got to be able to attack and defend at machine speed because that’s what the bad guys are doing across all of your attack surfaces.”

AI allows attackers to examine web applications, internet-facing infrastructure and other attack surfaces simultaneously rather than assigning people to investigate individual systems. The goal isn’t to automate vulnerability scanning but to automate how an attacker discovers weaknesses, determines whether they are exploitable, moves through the environment and establishes what business impact is possible.

“In many cases, customers are saying, ‘Look, the speed of human pen testers can’t keep up with algorithms, and I want to be able to do this more frequently,'” Hartley said. “And they’re saying to Horizon3, ‘Look, your test is as good or better than those humans I’ve used, but I want to use this at scale more frequently.'”