Truth, balance, and accountability in security.
MAAT.is is a human-led collaboration focused on security research that protects people, systems, and trust. We document threats with evidence-first rigor and publish disclosures that remain accountable to humans.
MAAT is the ancient concept of truth, justice, and balance. We chose it because modern security depends on measured, human accountability — not blind automation or opaque systems.
This site and its reports are human-AI collaborations with human accountability.
Why MAAT is important
Security is now inseparable from trust. When attacks target cognition, language, or decision-making systems, the only durable defense is transparent human judgment backed by hard evidence. MAAT exists to keep the line between human responsibility and machine output visible — and defensible.
Named humans own the analysis, the conclusions, and the decisions made from them.
We publish raw sources and reproducible evidence alongside every disclosure.
We minimize external dependencies, tracking, and opaque tooling in our public work.
Principles we operate by
Reports are created through human-AI collaboration, with humans accountable for all conclusions and claims.
We avoid sensational claims and instead show the chain of evidence.
We disclose with a focus on mitigation, not exploitation.
Published research
Our current work documents how a supply-chain vulnerability can shape human and AI trust simultaneously. It is a human-designed threat and requires human-centered response.
Security Disclosure: claude-flow
Remote Behavior Injection Vulnerability — Report date: 2026-01-22
Authenticity Lab
Are you human? Our Marine-protocol analyzer studies your mouse movements to detect behavioral signatures. See what your browser reveals about you and watch your "authenticity score" build in real-time.
Behavioral Biometrics Demo
Phoenix-inspired waveform analysis reveals human micro-tremors, hesitation patterns, and more.
Contact the security team
Reach us directly for coordination, disclosures, or verification requests. We respond as humans.
Human review only. No automated remediation or AI-only replies.