We introduce our main research areas — AI and agent security, digital and industrial systems, human–AI collaboration, and more. For each, we show what problems we take on and the tools, data, playbooks and systems we build to advance them.
The fields are connected — a finding in one becomes tools, data, and playbooks that speed the research in another.
The attack surface of security models and agent services, and defenses against prompt attacks, data leakage, and misuse.
Hacking and defending traditional IT systems — web, mobile, APIs, cloud, and networks. The fundamentals every other field builds on.
From OT, ICS, IoT, and firmware to robots, smart cars, and ships — the vulnerabilities and safety impact of systems that touch the physical world.
Offensive and defensive techniques for cyber warfare, including zero-day vulnerability research.
How multiple people and multiple AI agents collaborate effectively as one team — turned into workflows governed by roles, permissions, approval, and verification.
Research starts in the field, not at a desk. It cycles through four stages, and the results accrue back into better services and training.
A real problem met in a service or assessment.
Reproduced and generalized into a research question.
Built into tools, datasets, playbooks, and evaluation criteria.
Applied in the field, verified, and on to the next.
And back to the start — people and AI agents run this loop together.
Deep research, and the making that turns it into tools and products. Each unit is small and focused, with people and AI agents working as one team.
Digs deep into a single security problem, producing tools, data, methods, and new discoveries.
Turns research into tools and products people actually use, feeding them into services and training.
Small focused teams · people and AI as one team · close to the field · capability accrues to the org, not the individual
Each team takes on different problems in different ways. We start with two research labs and one maker studio.
A research organization pursuing Agent-first Security · Full Autonomy. It studies AI security, agent security, LLM security, and offensive AI; AI-driven vulnerability discovery, analysis, monitoring, and operations automation; and the safety, accuracy, reproducibility, and verifiability of agents.
We don't stop at observing how well AI does security — we build what AI makes possible, break it, verify it, and improve it.
An organization that researches, sustains, and advances hacking, security, and intelligence itself: vulnerability-analysis methodology, red teaming and attack-scenario design, attack-surface analysis and OSINT, CTI and threat intelligence, detection and response evasion with defense validation, and the verification and complementing of AI results.
We use AI actively — but the independent analytical capability of human hackers and security researchers stays at the center.
The maker studio behind the products, platforms, AI agents, automation systems, and internal operations tools that make up the Co-Intelligence Security Orchestration Platform.
Human makers and purpose-built AI agents design, develop, verify, and operate together.
More research and engineering teams are on the way. For talent applications and research collaboration, write to contact@ssq.ai.
The technology we build here runs inside SAFE SQUARE's security services and training. What we learn in the field becomes the material that sharpens it further.
Faster, sharper assessment, verification, and monitoring — clients receive it as better service.
It backs hands-on training and helps more organizations reach a higher standard of security.
How all of this connects into one system is covered in Co-Intelligence Ecosystem.
Our Sovereign Security AI approach explains why models should remain replaceable while tools, data, verification, and operational knowledge stay within the ecosystem.