Geordie AI raises $30m to find the AI agents companies cannot see

On 28 May, Geordie AI, a London-based company that sells security and governance software for AI agents, said it had closed a $30m (£22.4m) Series A round led by Balderton Capital. Crosspoint Capital joined, and existing investors General Catalyst and Ten Eleven Ventures followed on, taking total funding to $36.5m. James Wise, a partner at Balderton, is joining the board.

The round comes 13 months after the company was incorporated at Companies House, on 14 April 2025. Two of its three founders came from Darktrace, a well-known security software company. Henry Comfort, the chief executive, held senior roles at Darktrace, including US chief operating officer through its stock market listing and $5bn sale, according to Geordie’s website. Hanah Darley, chief AI and product officer, built Darktrace’s threat research function. Benji Weber, the chief technology officer, led engineering teams at Snyk and Mindgard.

The name is borrowed from the Geordie lamp, the miners’ safety lamp whose flame went out when gas built up underground. In the post that marked the company’s public launch and its seed round, co-led by General Catalyst and Ten Eleven Ventures, the founders wrote that mining lamps helped workers see risks “that they couldn’t see with their own eyes”. The software applies the same idea to agents: it shows which ones exist, what they can access, how they behave and what risks they create. A second product, Beam, works at runtime to shape and constrain what agents do.

The pitch rests on a narrow claim: companies deploying AI agents have lost track of how many they run, and the security tools they already own were not built to see them. Balderton’s round is a bet that this gap can carry a standalone company. The strongest evidence Geordie has published so far comes from customers who found more than they expected once they looked.

Customers who went looking

Owkin, an AI drug discovery and biomedical research company, runs hundreds of agents across more than 50 petabytes of data. A case study on Geordie’s website says Owkin found 327% more agents than it expected during a single proof of concept. According to the funding release, that trial averted exposure that Owkin’s own risk methodology put at $12m to $13m. “We’re seeing the iceberg that rocked the Titanic weeks in advance rather than the moment it appears on screen,” said Leo Cunningham, Owkin’s CISO.

Forge Holiday Group, a private equity-backed travel technology group with a portfolio of holiday letting brands, pushed AI across the business from the top. Jon Mattey, who leads cyber security at Forge, had concentrated on customer-facing AI such as booking assistants and chat. Geordie’s case study says the deployment turned his attention to agents on staff devices and in workflows that had never been inventoried. “When we deployed Geordie, what it actually showed me was I need to be thinking more about some of the stuff that’s emerging in the agentic space on the devices and with the workflows that are starting to crop up,” Mattey said.

Forge already reported cyber risk to its executive team in financial terms, comparing inherent with residual exposure. The case study says Geordie let Mattey extend that model to AI agents, and that the platform now covers more than 90% of Forge’s technology estate. When he raised concerns about AI workflows in Microsoft Power Automate, Geordie built a connector for them within weeks. Forge evaluated several emerging vendors in AI agent security before choosing Geordie, the case study says.

Growth on the company’s own numbers

Geordie says its annual recurring revenue grew 1,300% in the first five months of 2026. It has not published the revenue the percentage is measured against, which makes the figure a measure of pace rather than size. In March Geordie was named “Most Innovative Startup” at the RSAC 2026 Conference Innovation Sandbox in San Francisco, a contest in which 10 finalist security startups present to a panel of judges. Its website listed a team of 16 in May.

Mark Crane, a partner at General Catalyst, said the firm backed Geordie at seed when agentic AI security “was still a hypothesis”. He added: “In just six months, AI agents have moved from pilot to production faster than most enterprises were ready for.” Wise said Balderton had “followed the founders since their time at Darktrace and Snyk”.

What the Series A pays for

The company says the capital will go into the product and into larger engineering and go-to-market teams, with a focus on expanding its US operations. In the release Comfort framed the case around permission rather than prevention: “The organizations today that can safely approve and deploy AI agents are the ones that are capturing a new competitive advantage in their space.”

Geordie is selling to security teams that are being asked to approve agents rather than block them, and its two named case studies describe the same sequence: switch it on, find agents nobody had recorded, then put a money figure on the risk. Balderton’s $30m pays to repeat that sequence in the US. The 1,300% figure suggests buyers are starting to pay for it, though how much revenue sits behind the percentage is still for Geordie to disclose.