AI risk
AI security systems
How agents use tools, move meaning, expose data, and fail when people trust automation too much.
AI security research with practical outputs
Practical security research, tested in the field.
S6 studies AI agents, automation, SaaS control planes, SOC workflows, Splunk operations, and cyber harm. We publish what we can, test ideas in real systems, and turn useful findings into defensive tools.
current focus
Useful research, not theatre
active
Research that names the trade-offs
Tools built from real operational pain
Plain guidance for people under pressure
Private handling for capability that can be abused
plain version
We help people understand what is real, what is risky, and what is worth building next.
What S6 works on
Security problems where the details matter.
The work spans research, public guidance, controlled product builds, and private capability that needs careful handling.
Security operations
SOC and Splunk operations
Cleaner log data, better triage, source evidence, SC4S workflows, and fewer wasted analyst cycles.
Public good
Cyber harm reduction
Practical guidance, safer learning, and tooling that helps people reduce risk before recovery becomes painful.
Emerging threat research
Tracking the risks that are arriving before the playbooks catch up.
This is where S6 watches new attacker behaviour, AI-enabled failure modes, public-source exposure, and SOC coverage gaps. The output is plain research, defensive guidance, and careful handoff into products only when that is safe.
Read emerging researchResearch lane 01
AI and SaaS control-plane risk
How agents, plugins, OAuth scopes, workspace integrations, and SaaS automation can turn small trust mistakes into broad exposure.
Research lane 02
Exposed data and public-source leakage
Research into documents, metadata, misconfigured stores, and accidental public signals — handled with suppression and human review gates.
Research lane 03
SOC detection gaps
Emerging attacker behaviour mapped back to log-source quality, parser coverage, dashboard evidence, and what analysts can actually verify.
Research lane 04
Home and small-business cyber harm
Practical threat research for families, creators, and smaller teams who inherit enterprise-style risk without enterprise support.
Product stack
Public tools where possible. Private handling where needed.
SecHub, SC4S Manager, the SC4S Source Library, and cyber learning games can be discussed plainly. Sensitive OSINT and offensive automation stay invite-only or private because misuse risk is not theoretical.
View the product stack01
Research
Follow evidence before it becomes a tidy product claim.
02
Test
Build small proofs that show what a control can and cannot see.
03
Explain
Write the useful parts clearly enough for someone else to check.
04
Apply
Turn durable findings into better architectures, workflows, detections, and tools.
Start with the work
Read the research. Contact us when you need help applying it.
The useful conversations start with a field note, a lab, a technical question, or an operational constraint. Preferably more than one.