Hadrian, the Amsterdam-based offensive security startup, raised $40 million co-led by Forgepoint Capital International and SmartFin, bringing its total funding to $65 million2,3,4. The round lands days after Paris-based Fleuret AI closed a €4 million pre-seed led by RAISE Ventures5,6, extending a two-week funding wave that now exceeds $800 million across at least seven agent-security startups. The back-to-back rounds on both sides of the Atlantic confirm that agentic offensive security, using AI to attack systems before adversaries do, has become a distinct and well-capitalized investment thesis, not merely a feature inside broader cybersecurity platforms.
Armadin, the offensive-security startup founded by Mandiant creator Kevin Mandia, raised $255.5 million in a Series B at a valuation exceeding $2.5 billion, co-led by Andreessen Horowitz and Accel1,7,15. The round, announced October 1, brings Armadin's total funding to $445 million, only seven months after its public launch16. It lands in a week that also saw doxx.net raise $38 million17 and Reco close a $55 million Series B extension14, following Kontext's $4 million seed round and Island's $400 million round at a $6.4 billion valuation on September 2410,11,12,18.
Why it matters
Hadrian's CEO Rogier Fischer said the company's AI agents run more than 100 million attack simulations each day. Hadrian has raised $65 million, less than 6% of the $1.1 billion raised jointly by XBOW, Horizon3, and Armadin. That gap frames the competitive landscape: Armadin's $445 million total leads the offensive-testing category, while Hadrian and Fleuret AI are building from earlier stages in Europe, where no comparable player has yet reached growth-round scale.
Island's $6.4 billion valuation represents a $1.8 billion increase from its prior round9. Armadin's investor roster includes In-Q-Tel, the U.S. intelligence community's venture arm. In-Q-Tel's participation places Armadin's agent-swarm capability inside a national-security procurement pipeline, not just a commercial one.
The big picture
The offensive lane is filling fast. Armadin deploys hostile agent swarms: on August 3, the company said it had sent 26,000 AI agents to attack a live institution's network for three days, with the institution's consent, chaining 38 validated attack paths and producing 238 security findings. "Offense is uniquely advantaged right now. AI lets an attacker find and chain weaknesses faster than any human team can respond," Mandia said.
Hadrian's platform combines continuous exposure management with agentic penetration testing. Its Atlas product maps an organization's external attack surface, while Nova provides on-demand agentic penetration testing. Hadrian claims no other vendor offers both capabilities on one platform. Fleuret AI automates penetration testing through two AI agents, Emile and Champollion, which map a company's environment and attempt to exploit identified vulnerabilities.
ANALYSIS The market is splitting into two complementary lanes: offensive testing (Armadin, Hadrian, Fleuret AI) and runtime enforcement (Kontext, doxx.net, Reco). Both assume the same premise, that autonomous agents will act in ways their operators did not authorize, but they address it from opposite ends.
On the defensive side, Kontext's software sits between an agent and the systems it acts on, checking every requested action against security policy. Co-founder Jens Ernstberger framed the shift: "There are now three actors rather than two". Reco's $55 million extension brought its total funding to $140 million.
Between the lines
Island CEO Mike Fey told CNBC: "Every old control is breaking, so everything's up for grabs. I've never seen anything like it". Fleuret AI is targeting the gap between how quickly information systems evolve and how infrequently penetration tests are conducted; subsequent deployments can alter the attack surface shortly after an assessment has been completed. That framing, shared by Hadrian and Armadin, explains why all three offensive startups emphasize continuous or always-on testing rather than periodic audits: the agentic attack surface changes faster than any annual or quarterly review can capture.
Hadrian will use its funding to expand across EMEA and the U.S. while growing its engineering and research teams13. Fischer said most of the funds will go to research and development, with a significant portion allocated to U.S. expansion. Hadrian currently has 90 employees. Ernstberger said: "An AI agent can be properly authenticated, use an approved tool, and still take an action no one authorized". That architectural bet, intercepting at the tool-call level rather than the model layer, positions Kontext as model-agnostic middleware, a meaningful design choice as agent frameworks proliferate.
What's next
Armadin will spend its new capital on platform development, research, training, and customer deployment. Island plans to grow staff to 1,50019. Kontext's early customers cluster in financial services, a sector where compliance mandates could accelerate adoption of runtime controls. Hadrian's U.S. expansion puts it in direct competition with Armadin on American enterprise accounts. Fleuret AI's two-agent architecture targets European enterprises that update systems continuously but test security periodically. The regulatory backdrop adds pressure: Island's round arrives as the AI regulation debate intensifies in Washington.