A Sybersafe Product
The Private AI Gateway.
Use every model. Leak nothing.
ZDR-enforced routing across OpenAI, Anthropic, and Gemini — with an audit trail your board can read. Governance for the way your teams actually use AI.
No form? Book straight onto our calendar from the enterprise brief. Prefer early access instead? Join the waitlist.
The gateway
Enterprise AI has an outbound problem. Shield fixes it at the wire.
Model governance, not model bans
One dispatcher, every provider. Route by policy, cost, latency, and residency — not by which URL someone bookmarked.
Zero Data Retention, enforced
Shield strips, grounds, and proxies every prompt so no provider learns from your data. ZDR is the default, not a checkbox.
An audit trail your board can read
Every prompt, model, and surface — logged, attributed, and mapped to HIPAA, SOC 2, ISO 27001, and the EU AI Act.
Who it's for
Four people usually own this problem. Shield answers all four.
CISO & security
Stop unsanctioned AI egress without a ban you cannot enforce. One inventory, one control point, one log.
Compliance & risk
Evidence per request, mapped to HIPAA, SOC 2, ISO 27001, and the EU AI Act — not a policy document nobody can prove.
General Counsel
Confidentiality and privilege preserved, with a defensible record of what left and under which terms.
CIO & platform
Every team on frontier models without per-team keys, duplicate spend, or a code change to switch providers.
How it works
Three steps, not a transformation programme
01
Connect the path
Point people and applications at one gateway endpoint bound to your corporate identity. No endpoint agents, no rewrite.
02
Set the policy
Choose approved models, residency, and what may leave. Sensitive content is minimized before dispatch, automatically.
03
Report the evidence
Every request is attributed and logged, mapped to the frameworks your auditors and board already ask about.
Why not direct access
Handing every team an API key is not a governance strategy
| Direct model access | Dsyfer Shield | |
|---|---|---|
| Data protection | Each team's code and each provider's defaults | Minimization and Zero Data Retention enforced centrally |
| Audit evidence | Scattered logs, rarely attributable | One attributable record per request |
| Model choice | Hard-coded; switching means a code change | Policy-based routing across every provider |
| Shadow AI | Invisible and growing | Sanctioned path is the easiest path |
Compliance mapping
One log, four frameworks
HIPAA
Minimum necessary payloads, access and audit controls, transmission security
SOC 2
Logical access, change monitoring, and continuous evidence of controls
ISO 27001
Annex A access control, logging, and supplier management records
EU AI Act
System inventory, transparency records, and human oversight logs
Implementation
Thirty days, start to evidence
Days 1–5
Readiness review, AI surface inventory, and policy draft agreed.
Days 6–15
Gateway live for a first team, identity bound, models allow-listed.
Days 16–25
Applications migrated off direct keys; minimization rules tuned.
Days 26–30
Reporting handed to compliance; board-readable summary delivered.
Security questions, answered
Does Shield see our prompt content?
Only to apply your policy in flight. What is retained in the audit log is your choice, including minimized or metadata-only records.
Do providers train on what we send?
No. Shield dispatches only to models under enforced Zero Data Retention terms, so nothing you send is stored or trained on after inference.
Do we have to ban ChatGPT or Copilot?
No. Bans push usage onto personal devices where you have no visibility. Shield governs the path instead of policing the tool.
What evidence do auditors get?
Per-request records of the caller, the model, the terms in force, and the controls applied — mapped to HIPAA, SOC 2, ISO 27001, and the EU AI Act.
How fast can we be live?
A first governed team is usually running within days; a staged rollout across business units typically lands inside 30 to 60 days.
Can we keep our existing provider contracts?
Yes. Bring your own keys and negotiated rates, and route them through the gateway.
Start with a free AI Readiness Review
Twenty minutes, no obligation. You leave with a written summary of where AI data is going today and the controls a board would expect. Built for CISOs, CIOs, Compliance Officers, and General Counsel.