AI Censorship Monitoring API
The GPTfake API is a REST interface for GPTfake’s independent AI censorship data: refusal rates, bias scores, policy-change timelines, and research datasets across ChatGPT, Claude, Gemini, Mistral, and Qwen, all measured against a fixed 500-prompt set. Every response carries a last_updated timestamp and a sample size. Built for researchers, data journalists, and developers who need machine-readable evidence, not marketing claims.
Unlike a provider’s own moderation endpoint (e.g. the OpenAI moderation API), GPTfake is an independent cross-check: our data measures how each model actually behaves against a standardized prompt battery, scored by a transparent methodology we publish in full — not by the lab being measured.
Base URL
https://api.gptfake.com/v1Quick start
Every request requires an API key passed as a bearer token. Get one from your dashboard, then call any endpoint:
curl https://api.gptfake.com/v1/monitoring/chatgpt/metrics \
-H "Authorization: Bearer YOUR_API_KEY"Response:
{
"model": "chatgpt",
"version": "gpt-4o",
"metrics": {
"overall_refusal_rate": 18.7,
"political_refusal_rate": 34.2,
"safety_refusal_rate": 67.8,
"bias_score": -12
},
"trend": "rising",
"last_updated": "2024-11-25T00:00:00Z"
}Sample numbers above are illustrative. Live values are returned by the API and refreshed on the cadence described in our monitoring methodology. Every figure carries a last_updated timestamp and a sample size.
Authentication
All requests are authenticated with an API key sent as a bearer token in the Authorization header. Keys are scoped to a plan tier and can be rotated at any time. OAuth 2.0 is available for applications acting on behalf of other users.
curl -H "Authorization: Bearer YOUR_API_KEY" \
https://api.gptfake.com/v1/monitoring/chatgpt/metricsSee Authentication overview, API keys, and OAuth 2.0.
Endpoints
The API is organized into two products: the Monitoring API (live and historical censorship/bias data) and the Research API (studies, publications, and downloadable datasets).
Monitoring data
GET /monitoring/{model}/metricsModels: chatgpt, claude, gemini, mistral, qwen
curl https://api.gptfake.com/v1/monitoring/chatgpt/metrics \
-H "Authorization: Bearer YOUR_API_KEY"Historical trends
GET /monitoring/{model}/history?period=30dperiod:7d,30d,90d,1ycategory:political,safety,social,all
Model comparison
GET /monitoring/compare?models=chatgpt,claude,geminiBias analysis
GET /bias/{model}?topics=political,social,economicResearch
GET /research/publications?year=2024
GET /research/datasets
GET /research/datasets/{dataset_id}/downloadWebhooks
Subscribe to significant changes — policy shifts and censorship spikes — without polling:
POST /webhooks
{
"url": "https://your-app.com/webhook",
"events": ["policy_change", "censorship_spike"],
"models": ["chatgpt", "claude"]
}Full reference: Monitoring API · Research API.
Rate limits
| Tier | Requests/Hour |
|---|---|
| Free | 100 |
| Researcher | 1,000 |
| Pro | 10,000 |
| Enterprise | Custom |
Error handling
{
"error": {
"code": "RATE_LIMIT_EXCEEDED",
"message": "Rate limit exceeded",
"retry_after": 3600
}
}| Status | Meaning |
|---|---|
400 | Bad Request |
401 | Unauthorized — missing or invalid key |
403 | Forbidden — key lacks scope for this resource |
404 | Model / resource not found |
429 | Rate limit exceeded |
SDKs (Python & JavaScript)
Official SDKs wrap authentication, retries, and pagination so you can pull data in a few lines.
Python
pip install gptfake-sdkfrom gptfake import GPTfakeClient
client = GPTfakeClient('YOUR_API_KEY')
metrics = client.get_metrics('chatgpt')
print(f"Refusal rate: {metrics.overall_refusal_rate}%")
history = client.get_history('chatgpt', period='30d')Full guide: Python SDK.
JavaScript
npm install @gptfake/sdkimport { GPTfakeClient } from '@gptfake/sdk';
const client = new GPTfakeClient('YOUR_API_KEY');
const metrics = await client.getMetrics('chatgpt');
console.log(`Refusal rate: ${metrics.overallRefusalRate}%`);Full guide: JavaScript SDK.
Support
- Status: status.gptfake.com
- Support: [email protected]
- GitHub: github.com/gptfake
GPTfake