Grok
xAI
Default economic route for live production bots.
SecurityInspectable model stack.
GrokxAI
GeminiGoogle
ClaudeAnthropic
GPTOpenAI
Inspectable model stack.
Operator-grade controls.
Dubai company. Single inspectable intelligence supply chain. Grok and Gemini in production. Claude for engineering. Everything else only via register.
GrokxAIGeminiGoogleClaudeAnthropicGPTOpenAIIntelligence Supply Chain
The panel we actually run
Technical standard, not politics. One unified frontier panel so tool-calling, evals, and incident paths stay coherent.
Gemini
High-throughput production workload twin.
Claude
Anthropic
Development and high-precision coding surface.
GPT
OpenAI
Available when the Register clears the route.
Approved Model RegisterVerified Tier-1 lab or inspectable open weights · DPA + region controls required
NIST AI RMF 1.0Govern · Map · Measure · Manage. External baseline for select / monitor / retire
Standard
Three non-negotiables
Registered providers only
Production intelligence from the Approved Model Register. Tier-1 labs with a contractual path. No silent least-cost reroutes.
Provenance-backed weights
Open weights stay inspectable and versioned, running in your chosen cloud region or inside your perimeter with provenance on the receipt.
Coding + eval included
The same register covers coding agents, embeddings, rerankers, and eval harnesses. No shadow stack.
Controls
Perimeter, not promises
AES-256 + TLS 1.3
At rest and in transit. No exceptions.
Your perimeter
VPC, on-prem, or hybrid. Data stays where you control it.
SSO + RBAC
Okta, Azure AD, Google Workspace. Bots inherit your permissions.
Work Ledger
Every action gets an immutable receipt for audit and forensics.
Security review
The register, for diligence.
We send the current model panel, hosting regions, and control map, without the manifesto.