Program Overview
Anthropic runs a bug bounty program on HackerOne. The program has 14 in-scope assets and is managed by HackerOne's triage team.
14
In-Scope Assets
11h
Avg Response
98%
Efficiency
8d
Avg Bounty Time
37d
Avg Resolve
In-Scope Assets
| Asset | Type | Max Severity | Eligible |
|---|---|---|---|
| API & SDKs | AI_MODEL | Critical | Bounty |
| Claude Code | OTHER | Critical | Bounty |
| Claude Desktop Extensions and Claude.ai MCP servers | OTHER | Critical | Bounty |
| Claude in Chrome | OTHER | Critical | Bounty |
| Infrastructure & Internal Apps/Services | OTHER | Critical | Bounty |
| Leaked Employee API Keys | OTHER | Critical | Bounty |
| Official Clients | OTHER | Critical | Bounty |
| anthropic.atlassian.com | URL | Critical | Bounty |
| api.anthropic.com | URL | Critical | Bounty |
| claude.ai | URL | Critical | Bounty |
| console.anthropic.com | URL | Critical | Bounty |
| docs.anthropic.com | URL | Critical | Bounty |
| github.com/anthropics | SOURCE_CODE | Critical | Bounty |
| support.anthropic.com | URL | Critical | Bounty |
Out-of-Scope Assets
- https://github.com/modelcontextprotocol
Tips for Hacking Anthropic
- Read the policy — Understand what's in scope, out of scope, and any specific testing restrictions before you start.
- Enumerate the attack surface — Use subdomain enumeration and directory bruteforcing to map all accessible endpoints.
- Focus on high-impact bugs — Look for SQL injection, SSRF, and IDOR vulnerabilities first.
- Test authentication flows — Check for OAuth misconfigurations and CSRF in login/signup flows.
- Write clear reports — Include steps to reproduce, impact assessment, and suggested remediation. Use Burp Suite to capture evidence.
Frequently Asked Questions
How do I start hacking Anthropic?
Sign up on HackerOne, read the program policy carefully, review the in-scope assets listed above, and start testing. Always stay within scope and follow responsible disclosure guidelines.
Does Anthropic pay bounties?
Yes, Anthropic offers monetary rewards for valid security vulnerabilities.
What types of vulnerabilities does Anthropic accept?
Anthropic accepts reports for vulnerabilities found in their 14 in-scope assets. Common accepted vulnerability types include XSS, SQL injection, SSRF, IDOR, authentication bypass, and RCE. Check the program policy for specific exclusions.