Secure against and through AI
AI systems have now become so powerful that they can attack and infiltrate IT systems on their own. What once required expert knowledge and a great deal of manual work is now handled by agents that independently find vulnerabilities and play through entire attack chains. This capability is both a threat and an opportunity — depending on who uses it first.
Dangers of AI
State-of-the-art LLMs can now work fully autonomously for hours or even days, thanks to huge context windows. Sophisticated orchestrators let AI agents plan, form hypotheses, and try many different approaches in parallel. In doing so, the AIs don’t interact in the usual way — with a human through a chat interface — but via tool calling with digital tools; they write code and receive feedback from the connected systems. On top of that, AI capabilities are rising sharply, especially in IT and cybersecurity, so these systems can now carry out long attack chains and even identify new vulnerabilities.
Security through AI
Attacks on IT infrastructure are not a new risk. Since, given enough time and resources, the same attack methods can be carried out by humans, all critical IT systems must be tested regularly. Penetration tests are one of the most effective means of security assessment: IT systems are attacked on a trial basis to uncover vulnerabilities. However, they are time-consuming, expert-dependent, and hard to scale. The new capabilities of LLMs now make it possible to run such security tests largely automatically — protecting against both machine and human attackers.
Hand in hand
We believe AI speeds up testing but does not replace the judgment of experienced professionals. The best solution is collaboration with a human in the loop: the AI handles the (mass) execution, while humans control scope, strategy, and the remediation of every finding.
What now?
Even though IT attacks are nothing new, AI makes them easier and cheaper. The question is who sets the AI loose on a system first: attacker or defender. Those who test their own systems with AI-assisted pentests close gaps before someone exploits them. And because the world of AI is evolving at breakneck speed, a one-time test isn’t enough. What’s secure today may not be tomorrow, as new models and attack techniques emerge. Only those who test regularly stay secure.
We test IT and AI systems before attackers do. Our experts combine human judgment with autonomous AI agents to pinpoint exactly the vulnerabilities that become especially dangerous under AI-assisted attacks.
We draw on a broad selection of leading AI models and pick the most powerful tool for each task. Instead of one-off spot checks, we rely on regular, automated testing.
And we work exclusively with European infrastructure, fully GDPR-compliant, of course.
Services
- LLM pentesting of IT infrastructure
- AI-based vulnerability analysis of source code
- AI-based security monitoring
- Awareness training
- EU AI Act training
