01 · RÓLAM
Rácz-Akácosi Attila.
analytical + digital experience
first machine learning project
one team — no subcontractor chain
02 · POSITIONING
Turning innovation into strategic advantage.
As an expert team with more than two decades of experience in algorithmic systems and predictive analytics, we combine systems-level thinking with the latest AI security research.
As independent auditors, we proactively uncover hidden vulnerabilities in large language models (LLMs) through practical, real-world adversarial testing — helping Hungarian companies transform technological innovation from a business risk into a strategic advantage.
02 · PROFESSIONAL PHILOSOPHY
AI security is not just a technical problem.
The behavior of Large Language Models (LLMs) cannot be understood purely at the prompt level. It is the combination of three scientific disciplines that provides the framework through which we build security as a system-level strategy — not merely as an endless list of vulnerabilities.
01
Game theory
We model the strategic moves of attackers and defenders. An LLM is not static software — it is a live, adaptive interface where every defensive decision provokes an adversarial response. Anyone who views a red team audit purely through a technical lens misses why attackers will keep trying again and again.
02
Network science
We uncover the system’s invisible network of dependencies and critical points. A model’s vulnerability is rarely an isolated issue — it is more often a set of connections within a graph. For us, the question is not only “what broke,” but also “what did the chain reaction pull down with it.”
03
Chaos theory
It explains how the smallest error can lead to unpredictable, catastrophic consequences. A single-character hidden prompt injection can determine whether a chatbot remains a harmless customer support tool — or becomes a serious GDPR incident.
03 · KEY MILESTONES
Since 2007, deep inside digital systems.
From reverse engineering algorithms to machine learning, from predictive analytics to LLM red teaming. The technology changed — but the underlying logic governing complex systems has always remained the same.
2007
2009
2018
2024
2026
2007
Search Algorithms & Reverse Engineering
We began our journey by analyzing and reverse engineering algorithmic “black box” systems. We learned how to predict and influence the behavior of complex closed systems — and how to identify the gaps hidden behind their defensive layers.
2009
Digital Equality Conference
2018
2017 – 2018 | Predictive AI & E-Banking Summit
We applied machine learning in production environments using neural networks and predictive models — including financial and cryptocurrency datasets — and presented AI-driven forecasting results to executives of Hungarian financial institutions at the Budapest E-Banking Summit.
2024
Cybersecurity and AI in Practice
Through practical conference talks (WP Day) and major national media outlets (HVG, Origo, 24.hu, Bors), we educated the market about data-driven systems and the darker side of AI — including deepfake fraud, phishing, and AI-enabled manipulation techniques.
2026
2026 | aiq.hu — Independent LLM Red Team Audit
2026 | aiq.hu — Independent LLM Red Team Audit
04 · CONTACT
How can I help you?
Have an AI security question? Reach out — I’ll respond within 1 business day.