RedCell Labs
Research, technical reports, tools, and practitioner briefings are in preparation. We publish when we have something evidence-based to say — a discipline we recommend for AI systems as well.
In preparation
The pieces below are in active development. No fake publish dates, no back-dated thought leadership — they ship when they meet the bar.
Coming soon
Jailbreak resistance is one test class out of more than thirty. What a complete AI red team engagement actually covers — and why the gap between the two is where incidents live.
Coming soon
A practical question set for the approval meeting: identity, budgets, approval gates, logging, and the rollback plan. If the answers are vague, so is the risk acceptance.
Coming soon
Retrieval-augmented generation moves the security boundary into the document pipeline. How ingestion, embeddings, and access control decide what your assistant leaks.
Coming soon
Employees adopted AI before the policy existed. How to find shadow AI, understand the demand behind it, and channel it into governed alternatives that people actually use.
Coming soon
Three frameworks, three altitudes, one control map. How to align management system, risk process, and technical guidance without tripling the paperwork.
Coming soon
Model Context Protocol connects models to real systems with real permissions. The trust decisions hiding in tool definitions, and how to review them before someone else does.
Coming soon
The demo impressed the room; production inherits the liability. The engineering work between prototype and defensible system, itemized.
Coming soon
Containment looks different when the affected system holds credentials and keeps acting. Evidence, kill switches, and communication for AI incidents.
RedCell Labs briefings
A low-volume research briefing is planned alongside our first publications. Signup opens when there is something worth sending. Until then, email us and we will let you know.
If one of these topics is a live problem in your organization, you don't need to wait for the article. Email info@cyai.io and ask the question directly.