About Shadow Warden AI
Shadow Warden AI is a zero-trust AI security gateway. It sits between your application and every model it calls, screens each prompt before tokens are spent, strips secrets and personal data, and records only metadata — request content is never logged.
What it does
Filter
A nine-stage pipeline behind POST /filter: topology, obfuscation decoding, secret redaction, semantic rules, an embedding model, causal arbitration, reputation scoring, phishing checks and a final decision.
Redact
Fifteen secret and PII patterns plus an entropy scan for secrets no pattern knows about.
Gate agents
Session-level detection of injection chains across tool calls, so a multi-step agent cannot be walked into a harmful action.
Explain
Every decision carries a request id whose causal chain can be retrieved later as audit evidence.
It also hosts an experimental machine-to-machine marketplace for detection intelligence. That part runs on a test network; no real funds move.
Who runs it
Shadow Warden AI is independently built and maintained, with a public contact address and a published security-disclosure channel. The publisher's registered-entity details will be added here and to the privacy notice when incorporation completes; until then nothing on this site names a legal entity that does not yet exist.
What we do not claim
- Jailbreak detection is measured at 36.2% on our own 58-prompt adversarial corpus, with 0 of 35 benign prompts flagged. No higher figure is ours.
- We hold no certification of any kind. Control mappings are self-attested.
- There are no registered customers yet, so there are no customer references.
- Per-stage latency is not instrumented in production, so we quote none.
The full list lives in the trust centre.
Get in touch
- General, sales and support: vz@shadow-warden-ai.com — contact page
- Security disclosure: security.txt
- Privacy: privacy notice
- Developers: developer portal