AI operations
Agentic security systems
Multi-agent reasoning, autonomous investigation loops, controlled tool use, and the hard boundary between useful autonomy and unsafe automation.
Research-led security for emergent AI systems
Exploring the security edge of emergent AI tech.
S6 Security Labs studies how agentic AI, automation, SaaS control planes, and security operations collide in the real world. We publish research, build public labs, test uncomfortable assumptions, and turn useful findings into practical defensive capability.
research.s6securitylabs.com
Live research stream
online
Public experiments and defensive labs
Security economics and tool trade-offs
AI control-plane risk and data movement
Detection engineering, DLP, and SOC workflow
current focus
AI workflows that can retrieve, reason, act, and leak meaning without moving data in familiar shapes.
Research first
Less vendor copy. More evidence.
The front page should not read like a catalogue. S6 is a lab: we investigate what is changing, publish what we can, and keep the work grounded in tests, threat models, demos, and operational consequences.
Security operations
Defensive AI and SOC economics
How small teams can use AI to compress triage, detection engineering, reporting, and log investigation without surrendering judgement to a black box.
Research notes
Emergent platform risk
AI apps, connectors, MCP servers, SaaS control planes, semantic data movement, and the new failure modes showing up around modern enterprise AI.
Public research hub
The research site is now the front door for experiments.
Research notes, defensive labs, demonstrations, and exploratory work live at the dedicated research site. The main S6 site now points readers there before it pushes them into product pages.
Open research.s6securitylabs.com01
Explore
Track early technical changes before they become tidy product categories.
02
Test
Build small labs and proofs that show what a control can and cannot see.
03
Publish
Write up the useful parts with enough source material and evidence to be checked.
04
Apply
Turn durable findings into better architectures, workflows, detections, and tools.
Start with the work
Read the research. Then talk to us if it maps to a real problem.
The useful conversations start with evidence: a field note, a lab, a technical question, or an operational constraint. Preferably more than one.