Sinnlabs is the R&D wing of Saatvix. We research the questions that don't have off-the-shelf answers: how AI agents behave when no one is watching, how security policies conflict in practice, and whether human cyber risk can be measured like credit risk.
Each research area addresses a gap in how organizations currently think about security. These aren't theoretical exercises. They connect directly to our consulting practice.
AI systems are being deployed faster than organizations can monitor them. Traditional security tools detect network intrusions and malware. They don't detect an AI agent that gradually escalates its own permissions, rephrases blocked requests until they succeed, or switches languages to bypass content filters. We're building detection methods that focus on behavioral signals rather than signature matching.
Every organization with more than a handful of policies has conflicts buried in them. A data retention rule that contradicts a privacy deletion requirement. An access control policy that blocks a compliance reporting obligation. These conflicts are invisible until they cause an incident. We apply formal methods from theoretical computer science to detect and resolve these conflicts before they surface as real-world failures.
Credit scores changed lending by giving institutions a standardized way to quantify financial risk. Cybersecurity has no equivalent for human risk. We're researching a behavioral risk scoring model that gives organizations a consistent, evolving measure of their human-layer exposure. Not a one-time phishing test. A continuous, contextual score.
Sinnlabs research is grounded in established academic disciplines, not trend-chasing.
Our policy conflict resolution research uses formal reasoning methods developed in academic computer science. We treat security policy sets as structured logical systems, not text documents, and apply conflict detection algorithms to surface contradictions before they cause incidents.
Our AI detection research focuses on behavioral patterns, not signature databases. We detect patterns in how AI systems respond to constraints over time. Our approach catches systematic evasion techniques that keyword and signature-based filtering cannot.
Sinnlabs builds on open-source infrastructure. Our research tools integrate with Wazuh, OPA, and standard security pipelines. We believe defensibility comes from the method, not from locking customers into proprietary platforms.
We actively seek academic collaborators, security researchers, and organizations interested in co-developing these tools. If you're working on related problems, we'd like to hear from you.
Our three research themes are designed to work together. Each one solves a piece of a larger puzzle.
Behavioral AI monitoring identifies anomalous patterns in real time.
Policy conflict resolution ensures your response rules don't contradict each other.
Human risk quantification gives you a continuous measure of organizational exposure.
We're looking for research collaborators, early-access partners, and people who want to join the lab.