# Team\-wide Zero Data Retention and prompt training controls now on AI Gateway

**Published:** April 6, 2026 | **Authors:** Matt Lenhard, Jeremy Philemon, Walter Korman, Jerilyn Zheng

---

AI Gateway now supports Zero Data Retention (ZDR) at the team level, removing the need to configure opt-outs or reach agreements with each provider individually. It routes requests only to providers where ZDR agreements are in place, with support for Anthropic, OpenAI, Google, and many more models.

**Three compliance controls are now available:**

### **Team-wide ZDR**

Enable team-wide ZDR from the [AI Gateway Dashboard Settings](https://vercel.com/d?to=%2F%5Bteam%5D%2F%7E%2Fai-gateway%2Fsettings&title=AI+Gateway+Settings) to enforce Zero Data Retention on every request your team makes, with no code changes required. Requests are routed to ZDR-compliant providers that have agreements with Vercel AI Gateway.

### **Request-level ZDR & No Prompt Training**

```typescript
import type { GatewayProviderOptions } from '@ai-sdk/gateway';
import { streamText } from 'ai';
const result = streamText({
  model: 'anthropic/claude-sonnet-4.6',
  prompt: 'Summarize this internal document.',
  providerOptions: {
    gateway: {
      disallowPromptTraining: true,
      zeroDataRetention: true,
    } satisfies GatewayProviderOptions,
  },
});
```

**ZDR**

You can also configure ZDR on the request level by setting `zeroDataRetention: true` per request. This restriction also does not apply to BYOK requests.

**Disallow Prompt Training**

Prevent providers from using your prompt data to train their models by setting `disallowPromptTraining: true` per request. Enabling Zero Data Retention automatically includes training opt-out, so there's no need to configure both separately.

**Pricing**

| **Option** | **Cost** | **Availability** |
| Team-wide ZDR | $0.10 per 1,000 requests | Pro and Enterprise |
| Per-request ZDR | Free | Pro and Enterprise |
| Disallow Prompt Training (per-request) | Free | All users |

Team-wide and per-request settings work together. If either is enabled, ZDR is enforced. Each response includes metadata showing which providers were considered and which were filtered out, giving you an audit trail of policy enforcement.

These controls work with the AI SDK, Chat Completions API, Responses API, Anthropic Messages API, and OpenResponses API.

Read the [Zero Data Retention documentation](https://vercel.com/docs/ai-gateway/capabilities/zdr) and [No Prompt Training documentation](https://vercel.com/docs/ai-gateway/capabilities/disallow-prompt-training) for more details.

---

📚 **More updates:** [View all changelog entries](/changelog/sitemap.md) | [Blog](/blog/sitemap.md)