This Milan Startup Raised $3.4M to Trap AI-Powered Hackers With AI Decoys

Jan de Vries · 2026-07-27

Cybersecurity is a nonstop battle, and one that humans can no longer fight alone. That's the blunt assessment from Mario Candela, CEO and founder of Beelzebub, a Milan-based AI-native cybersecurity platform that just announced a $3.4 million Seed funding round. The investment was led exclusively by Italian DeepTech VC United Ventures. ### What Beelzebub Actually Does Here's the thing about modern cyberattacks: they're not just faster than human teams can handle — they're smarter. Beelzebub doesn't try to build a higher wall. Instead, it assumes the wall is already breached. The platform plants AI-powered traps that look like real infrastructure — servers, databases, APIs, cloud systems — all over a company's network. When an attacker touches one, the system detects it, blocks the attack, isolates the machine, and initiates response actions before any damage can occur. Think of it like a smart home security system that doesn't just lock the door but sets up fake rooms filled with tripwires. The moment an intruder steps into one, the system knows exactly where they are and what they're after. ### The Funding Details This $3.4 million Seed round follows a $343,000 pre-Seed from strategic investors and advisers, bringing Beelzebub's total funding to about $3.77 million. The company plans to use the capital for three things: - **Expand the research team** — more brains building better traps - **Open new commercial offices** in Rome and San Francisco by the end of the year - **Acquire clients across Europe**, with a special focus on organizations regulated by NIS2, the EU's updated cybersecurity directive Beyond that, Beelzebub is already building new tech to protect AI agents directly — because if you're using AI to run your business, attackers will target those agents too. ### How It Works: The Three-Pronged Platform Beelzebub's platform has three main components, and they work together in a continuous loop: **Arcangelo** handles the red team (attack) side. It simulates targeted attacks against your own network to find weak spots before real attackers do. **Beelzebub Managed** covers the blue team (defense). It plants those LLM-powered traps throughout your infrastructure. The traps update in real time to mimic new CVEs (Common Vulnerabilities and Exposures), which forces any dormant attackers inside your network to reveal themselves. **Caronte** is the AI analyst. After a catch-and-kill operation, Caronte reverse-engineers the malware and delivers a full incident report in a fraction of the time a human analyst would need. It can run entirely on-premise — essential for enterprises and government organizations that can't risk sending sensitive data to cloud-based malware analysis tools. > "The cybersecurity market is undergoing a structural reset: when offensive AI scales faster than any human team can respond, the only viable defense is one that operates at the same speed and intelligence," says Massimiliano Magrini, co-founder and Managing Partner at United Ventures. ### Why This Matters Now Beelzebub started as Mario Candela's research project in Milan back in 2021. It was incorporated in July 2025. The platform is built as a SaaS or on-premises solution, with more than 60 independent researchers feeding live threat intelligence into the system as new attack methods emerge. It's NIS2-ready by design, which is a big deal for European companies facing stricter cybersecurity rules. What makes this approach different is the "assumed breach" mindset. Most cybersecurity tools act as perimeter defense — they try to keep attackers out. Beelzebub assumes attackers are already inside, waiting for the perfect moment to strike. The moment they make a move, the traps close. For companies that can't afford to compromise — and in 2025, that's pretty much everyone — this kind of adaptive, AI-driven defense isn't just clever. It's becoming essential.