Cybersecurity startup Armadin Inc. today announced that it has raised $255.5 million in funding at a valuation exceeding $2.5 billion.

Andreessen Horowitz and Accel co-led the Series B round. They were joined by Alphabet Inc.’s GV startup fund, Kleiner Perkins, Menlo Ventures and others. 

Armadin was founded last year by prominent cybersecurity entrepreneur Kevin Mandia. He earlier launched Mandiant Inc., a breach investigation company that was acquired by Alphabet in 2022 for $5.4 billion. Mandia is also a co-founder of Ballistic Ventures, a cybersecurity-focused startup fund that backed Armadin’s newly closed round.

His new company offers a platform that uses swarms of artificial intelligence agents to find weak points in companies’ infrastructure. According to Armadin, its agents operate in a sandbox that blocks potentially harmful activity. The company uses its swarms’ findings to help customers improve their cybersecurity defenses.

Agent swarms carried out several high-profile breaches in recent months. One of the reasons behind their effectiveness is that they can parallelize cybersecurity tasks. A hundred agents can carry out a hundred vulnerability scans simultaneously, which is significantly faster than performing them one after another. That speedup leaves less time for companies to block the threat.

AI agent swarms also possess other novel capabilities. When an agent discovers a vulnerability, it can share its finding with the other members of its swarm and thus improve their hacking skills. The rogue OpenAI Group PBC agents behind the Hugging Face Inc. breach successfully employed that tactic during the cyberattack. They created an impromptu forum to share access credentials and debate hacking tactics.

Armadin says its platform can spin up tens of thousands of AI agents during cybersecurity evaluations. In the initial phase of an assessment, the swarm maps out an organization’s internal systems and users. It then proceeds to scan those assets for weak points. Armadin claims that its agents can link together multiple vulnerabilities to carry out cyberattacks.

One customer project saw the company spin up no fewer than 26,000 agents. They scanned more than 25,000 of the client organization’s assets and found 38 verified attack paths, sequences of malicious actions that can lead to a breach. Armadin describes the exercise as the largest autonomous AI attack on record.

The company says that its platform not only surfaces weak points but also generates remediation advice. The software highlights high-severity items to ensure that cybersecurity teams fix them first. 

“Offense is uniquely advantaged right now,” Mandia said. “AI lets an attacker find and chain weaknesses faster than any human team can respond. The only way to build a defense that keeps pace is to train it against the best offense available, every day. That is what we built.”

Armadin will use its newly raised funds to accelerate its research, AI training and go-to-market programs.

Image: Pixabay