AN
HackerOneOffers Rewards
Anthropic
Anthropic appears to run a bug bounty or vulnerability disclosure program on HackerOne. It also offers rewards for valid security reports.
Created: May 11, 2026
Last Updated: October 5, 2026
In Scope (14)
API & SDKs ( [Core] Anthropic's API enables enterprise customers to programmatically interact with Claude, supporting their diverse business requirements. Claude is a large language model (LLM) that has been trained to be a helpful assistant in a conversational tone. For additional information on the API and SDKs, please see the following documentation: [API](https://docs.anthropic.com/claude/reference/getting-started-with-the-api/) (Note: Please remember to use the [staging hostname](https://api-staging.anthropic.com/).)
[Python SDK](https://github.com/anthropics/anthropic-sdk-python)
[Python Agent SDK](https://github.com/anthropics/claude-agent-sdk-python)
[Typescript SDK](https://github.com/anthropics/anthropic-sdk-typescript)
[Typescript Agent SDK](https://github.com/anthropics/claude-agent-sdk-typescript) )
Claude Code ( [Core] **In Scope:**
* Bypassing permission prompts for unauthorized command execution (excluding pre-approved or user-allowed commands)
* Bypassing permission prompts for file write operations outside working directory
* Misrepresenting parameters or tools in permission prompts by displaying different information than what will actually be executed (see exclusion below related environment-specific settings).
* Executing commands or tools invisibly to users
* https://github.com/anthropics/claude-code
* https://github.com/anthropics/claude-code-action **Out of Scope:**
* Abusing intended functionality of Claude CLI * Using aliased commands, symlinks or other environment-specific settings to bypass permission prompts
* Local storage of Claude Code credentials, configuration and logs )
Claude Desktop Extensions and Claude.ai MCP servers ( [Non-Core] Connectors (e.g., desktop extensions, local MCP servers, and remote MCP servers) installable from the Claude.ai & Desktop Connectors directory may be developed by Anthropic or created independently by third-party developers. Vulnerabilities discovered in Anthropic developed software are in-scope for Anthropic's HackerOne program. Note that any vulnerabilities discovered in software developed by third parties (whether or not they are included in Anthropic's Connectors Directory) are not in scope for Anthropic’s HackerOne program and should be reported directly to those third-party developers. For details on reporting, see the extension page in Claude Desktop. We ask that vulnerability reports regarding third-party connectors are only shared with Anthropic if a satisfactory response is not received from the responsible developer. )
Claude in Chrome ( [Non-Core] https://chromewebstore.google.com/detail/claude/fcoeoabgfenejglbffodgkkbkcdhcgfn )
Infrastructure & Internal Apps/Services ( [Non-Core] In scope: - Vulnerabilities in Anthropic infrastructure that could lead to unauthorized access, data exposure, or service disruption. Examples include server misconfigurations, exposed internal services, subdomain takeovers, etc.
- Vulnerabilities in internal Anthropic web applications and services that are not intended for public access. This includes things like admin panels, support portals, build systems, monitoring tools, etc. Issues could include authentication bypasses, privilege escalation, XSS, etc.
- Vulnerabilities leading to unauthorized disclosure of confidential Anthropic information through third-party SaaS/Cloud platforms used by Anthropic. This includes unintended public exposure of documents, data, source code, credentials or other sensitive information. Please note: - This is not to be construed as an authorization to conduct security testing against the infrastructure of SaaS/Cloud vendors, please check with t
Leaked Employee API Keys ( [Non-Core] API keys belonging to employees that are publicly leaked. Hackers should submit the leaked credentials to the program and should NOT test their validity beyond authenticating and then immediately deauthenticating - without exercising any functionality. We will test validity and measure impact. At Anthropic's discretion in cases of critical impact, this may be treated as a Core asset. )
Official Clients
anthropic.atlassian.com ( [Non-Core] Note: It is working as intended that there are 4 queues accessible with an external account: AP Inbox, Finsys Intake, and Dev Accounts Payable. Reports about this will be closed as Informative. If you are able to access any sensitive information, please do open a report. )
api.anthropic.com ( [Core] )
claude.ai ( [Core] Claude, a next-generation AI assistant trained to be helpful, honest and harmless, is accessible through a chat interface on claude.ai. Claude is capable of a wide variety of conversational and text and image processing tasks and can help with use cases including summarization, search, creative and collaborative writing, Q&A, coding, and more. Claude can be accessed via [Claude.ai](https://www.claude.ai), which contains features such as 100k context windows and image & file uploads. )
console.anthropic.com ( [Core] )
docs.anthropic.com ( [Non-Core] )
github.com/anthropics ( [Non-Core] Note: Vulnerabilities in repos that are marked as archived or repos that are forks of other open source projects are excluded from our bug bounty unless the issue can be demonstrated to impact Anthropic assets. )
support.anthropic.com ( [Non-Core] )
Out of Scope (1)
https://github.com/modelcontextprotocol ( Issues in OSS Model Context Protocol code should be reported directly to MCP maintainers via the Security page for the affected repo. Issues affecting implementations of MCP in Anthropic products and services (on claude.ai, Claude Desktop, etc) are still in scope and should be submitted under the affected asset. )