The Terms of Service confusion that's fooling thousands of developers—and the simple fix

## The Controversy That Isn't (Mostly)

When a Hacker News user posted ["Brave Search API forbids use with AI agents"](https://news.ycombinator.com/item?id=46822822) earlier today, it sparked a question that's been brewing since January: Are tools like OpenClaw violating Brave's Terms of Service?

The short answer: **No—if you pick the right plan.**

But the confusion is real, and it's Brave's fault for poor UX during signup. Let me explain.

## The ToS Warning That Scared Everyone

When you sign up for Brave Search API, you see this warning:

> "By subscribing to this plan, you agree to abide by the Terms of Use. **These Terms of Use prohibit using responses for AI inference.**"

Developers rightfully panicked. If AI inference is prohibited, how can Claude, OpenClaw, or any LLM-powered agent use Brave Search?

The answer was hiding in a dropdown most users never changed.

## The Two Plan Types (That Look Identical)

Brave offers **two categories** of API plans, each with Free/Base/Pro tiers:

| Category | AI Inference | Use Case |
| --- | --- | --- |
| **Data for Search** | ❌ Prohibited | Traditional search UIs |
| **Data for AI** | ✅ Explicitly Permitted | [AI agent](/content/articles/clawdbot-mac-mini-setup-guide/index.html)s, LLMs, RAG |

**The problem**: The signup flow defaults to "Data for Search," which shows the scary ToS warning. Most developers never realize there's a "Data for AI" option that explicitly allows what they need.

Here's what the [Brave Search API pricing page](https://brave.com/search/api/) actually says:

- **Free AI**: "Get familiar with the API" — 2,000 queries/month
- **Base AI**: $5/1,000 requests — "Rights to use in AI apps"
- **Pro AI**: $9/1,000 requests — "Rights to use in AI apps"

The "AI" suffix matters. It's not just branding—it's a different license.

## The Official Clarification (From Brave Directly)

On January 17, 2025, Jonathan Sampson from Brave [clarified on GitHub](https://github.com/modelcontextprotocol/servers/issues/522#issuecomment-2598379335):

> "Great question. If you check brave.com/search/api/, you'll find that we offer 2 free account types. The first is **Data for Search**, where AI inference **is not** permitted. The second is **Data for AI**, where AI inference is **explicitly permitted**."

He later added context about what each plan allows:

- **Data for Search**: Display search results to users. No AI processing.
- **Data for AI**: Feed results into LLMs, RAG pipelines, AI agents. ✅
- **Data with Storage Rights**: Store results, train models. (Enterprise tier)

## The Remaining Gray Area

Even with the "Data for AI" plan, some restrictions remain:

From the Terms of Service:

> "2.b.i. Licensee shall not... **store the results of the API or any derivative works from the results**"

> "2.b.xiv. Licensee shall not... **redistribute, resell, or sublicense the Content**"

This raised follow-up questions on the GitHub thread:

**Developer concern**: "If my SaaS performs searches with the API, processes results using an LLM, and stores them in a DB—am I breaching the ToS?"

**Brave's response** (paraphrased): The Data for AI plan allows inference and presenting results to users. Storing raw API responses requires the "Storage Rights" tier. The restriction is about caching/reselling the structured API response, not about using search results in your application.

**The nuance:** Showing an AI-generated answer that used search results = ✅ allowed. Storing the raw API JSON for reuse = ❌ requires Storage Rights plan.

## What This Means for OpenClaw Users

**If you're using OpenClaw with Brave Search:**

1. Go to [brave.com/search/api](https://brave.com/search/api/)
2. Select **"Data for AI" plans** (not "Data for Search")
3. Subscribe to Free AI, Base AI, or Pro AI

That's it. You're compliant.

**Current setup command:**

```
openclaw configure --section web
```

When prompted, enter your API key from a "Data for AI" plan.

## Comparison: How Other Search APIs Handle This

| Provider | AI Usage | Notes |
| --- | --- | --- |
| **Brave (Data for AI)** | ✅ Free tier + paid | Must select correct plan type |
| **Perplexity Sonar** | ✅ All tiers | Built for AI, implicit licensing |
| **Google Custom Search** | ⚠️ Gray zone | ToS prohibits "automated queries" |
| **SerpAPI/Serper** | ❌ Legal risk | Scrapes Google/Bing |
| **Tavily** | ✅ All tiers | Purpose-built for AI agents |

## TL;DR

1. **The controversy is a UX problem**, not a legal one
2. **Brave offers "Data for AI" plans** that explicitly allow AI inference
3. **Most developers signed up for the wrong plan** because the UI is confusing
4. **Fix it**: Switch to Free AI / Base AI / Pro AI tier
5. **OpenClaw + Brave Search = compliant** when configured correctly

## Sources

- [Hacker News Discussion](https://news.ycombinator.com/item?id=46822822) — January 30, 2026
- [GitHub Issue #522: Clarify possible Brave Search TOS violation](https://github.com/modelcontextprotocol/servers/issues/522) — January 2025
- [Brave Search API Pricing](https://brave.com/search/api/)
- [Brave Blog: Zero Data Retention](https://brave.com/blog/search-api-zero-data-retention/)
- [OpenClaw Docs: Web Tools](https://docs.openclaw.ai/tools/web)