Armadin Raises $255.5 Million to Scale AI Attacks That Test Company Defenses

Armadin has secured $255.5 million to scale a security platform built to attack company systems before adversaries do. The October 1, 2026, Series B, co-led by Andreessen Horowitz and Accel, values the startup above $2.5 billion. Its funding announcement backs a specific proposition: AI agents should continuously find routes into an organization, not just catalog individual weaknesses.

The company is led by Kevin Mandia, the founder of Mandiant, which Google bought for $5.4 billion in 2022. His latest business shifts the testing work toward always-on groups of AI agents. Rather than organize a single human-led attempt to break in, Armadin describes software that keeps probing defenses and connects the weaknesses it finds.

The route matters more than the individual flaw

Armadin says its platform uses a swarm—a group of specialized AI agents—to reason across an organization's exposed systems. The agents chain weaknesses into validated attack paths. A weakness judged low-severity on its own can become part of a larger route to compromise when connected with others. That connection is the core of the company's product pitch.

The release describes paths that start with an outsider executing code at the network's perimeter without signing in. From there, an attack can move across systems and reach a full cloud compromise. Armadin says it shows security teams both the route an attacker could take and the scope of the damage, so they can break that route before exploitation.

That differs from the two approaches Armadin criticizes in its announcement. Periodic penetration tests are scheduled efforts to break into systems and report weaknesses. Scanners, meanwhile, score findings individually. Armadin argues that the first approach is too infrequent for AI-driven attacks, while the second misses how separate findings fit together. Those are the company's arguments, not independent comparative test results.

Capital for the platform and the selling effort

Bain Capital Ventures and Redpoint joined as new investors. Returning participants included 8VC, Ballistic Ventures, Google Ventures, In-Q-Tel, Kleiner Perkins and Menlo Ventures. Alongside the two co-leads, that roster funds Armadin's stated plan to scale its security platform, research and training, as well as its go-to-market work—the effort to reach and sell to customers.

The investor thesis is explicit. Accel partner Ping Li argues that AI-powered offense is the best defense against agents used by adversaries. Andreessen Horowitz general partner David George points to Mandia's breach-response experience and a team combining offensive-security specialists with AI engineers.

Production use, without named customers

Armadin says it is already running attack campaigns in production for Fortune 500 enterprises and government customers. That places its claimed deployments in live operating environments, rather than describing only a future product ambition.