Beelzebub, a Milan-based AI-native cybersecurity platform built to defend digital assets against AI-driven cyber attacks, has closed a €3m Seed round as offensive AI outpaces human defenders.
According to EU Startups, the round was led solely by United Ventures, the Italian DeepTech venture capital firm. It follows a €300k pre-Seed backed by strategic investors and advisers, taking the startup’s overall funding to €3.3m.
The fresh capital will go towards growing the research team, opening commercial offices in Rome and San Francisco before the year is out, and winning clients across Europe, with a particular focus on organisations covered by the NIS2 directive. The firm has also started developing new technology aimed at protecting AI agents themselves.
Beelzebub started life as a research project by founder Mario Candela in Milan in 2021 before being incorporated as a company in July 2025. Its platform is designed around the principle of assumed breach rather than perimeter defence.
In the company’s view, the wisest assumption is not that an intruder may break in at some unknown future date, but that one is already sitting inside the network awaiting the right moment to act. When that embedded attacker moves, the platform detects and ensnares them within a closed loop of attack simulation, deception and automated threat analysis.
The technology can be deployed as SaaS or on-premises for highly sensitive environments, including LLM inference running on hardware supplied by Beelzebub, and is NIS2-ready by design. Over 60 independent researchers worldwide contribute live threat intelligence as new attack techniques surface.
The platform operates a purple team structure. Arcangelo handles the red team side by simulating targeted attacks, while Beelzebub Managed takes the blue team role, seeding a company’s infrastructure with LLM-powered decoys that imitate genuine assets such as servers, databases, APIs and cloud systems.
Once an intruder interacts with a decoy, the system blocks the attack before damage occurs, isolates the compromised machine and triggers response actions. Decoys refresh in real time to mimic newly published CVEs, flushing out dormant attackers hiding in the network. Because the two products form a single automated offence-defence loop, the startup says there are no handoffs between tools or vendors.
Once a threat has been neutralised, Caronte, the platform’s AI analyst, reverse-engineers the malware and produces a complete incident report far faster than a human analyst could. Caronte can also operate fully on-premise, a requirement for enterprises and government bodies unable to send sensitive data to cloud-based malware analysis services.
Beelzebub CEO and founder Mario Candela said, “Cybersecurity is a nonstop battle, and one that humans can no longer fight alone. AI-powered attackers are too powerful, too competent, too fast, and too numerous, meaning the only way to fight back is at the same pace and intensity. Beelzebub’s product adapts to new types of malware and is always updated to match the current state of the most sophisticated attacks. This seed round will supercharge our efforts to bring modern cybersecurity to the companies who cannot afford to compromise.”
Copyright © 2026 RegTech Analyst
Copyright © 2018 RegTech Analyst
