Model Allowlist — GPT-4 + Claude 3

​

Policy JSON

{
  "registry": {
    "models": ["gpt-4o", "gpt-4o-mini", "claude-3-sonnet", "claude-3-haiku", "claude-3-5-sonnet"],
    "strict": true
  }
}

policy = {
    "registry": {
        "models": [
            "gpt-4o",
            "gpt-4o-mini",
            "claude-3-sonnet",
            "claude-3-haiku",
            "claude-3-5-sonnet",
        ],
        "strict": True,
    }
}

​

What It Does

• Maintains an allowlist of 5 approved models across OpenAI and Anthropic
• Denies requests for any model not on the list (e.g., gpt-3.5-turbo, claude-3-opus)
• Strict mode means no fallback — if the requested model isn’t allowed, the request fails
• Prevents agents from autonomously selecting unapproved, untested, or expensive models

​

Example

{
  "action": "DENY",
  "risk_score": 0.7,
  "reasons": ["model_not_in_allowlist"],
  "evidence": {
    "requested_model": "claude-3-opus",
    "allowed_models": ["gpt-4o", "gpt-4o-mini", "claude-3-sonnet", "claude-3-haiku", "claude-3-5-sonnet"],
    "strict": true
  }
}

​

Install

# Install via CLI
tealtiger policy install model-allowlist-gpt4-claude3

# Or copy the JSON above into your tealtiger.config.json

​

Customize

​

Related

• Tool Allowlist — Basic — Control tool access alongside model access
• Provenance Verification — Verify model versions and integrity
• Model Tier Enforcement — Cost-based model restrictions
• TealRegistry Documentation — Full API reference