OpenClaw (formerly Clawdbot and Moltbot) is an open-source personal AI agent that runs a local gateway on the user's machine and acts through the user's own model providers, messaging apps and tools. It ships an official macOS menu bar app, a native Windows Hub app and Linux packages, plus a one-command terminal install. Not to be confused with lookalike sites such as clawdbotai.online.
[ToS] explicitly states: 'We do not use your messages to train our own models.' This clear, unambiguous commitment prote
No training on user data means no opt-out is necessary.
Low risk across data training, licensing, and sharing dimensions.
[ToS] explicitly states: 'We do not use your messages to train our own models.' This clear, unambiguous commitment protects company data from being used to train OpenClaw's AI models. The commitment applies to the service as described.
“[ToS] 'We do not use your messages to train our own models, we do not sell or rent your data, and we do not run advertising or third-party analytics inside the bots.'”
[ToS] states 'You keep any rights you have in what you write' and grants permission limited to 'process your messages as described in the Privacy Policy' for 'the purpose of operating the service.' This restricts the license to service operation only, with no sublicensing or perpetual rights claimed.
“[ToS] 'You keep any rights you have in what you write. You grant us permission to process your messages as described in the Privacy Policy below, for the purpose of operating the service.'”
[ToS] discloses that 'AI assistant bots send message text and relevant context to third-party AI model providers, which currently include OpenAI and Anthropic' but states this 'Provider handling is governed by their own terms and privacy policies.' While essential sharing is disclosed, the documents lack explicit Data Processing Agreements or published sub-processor lists, creating moderate concern about control over third-party handling.
“[ToS] 'To produce a reply, AI assistant bots send message text and relevant context to third-party AI model providers, which currently include OpenAI and Anthropic. Provider handling is governed by their own terms and privacy policies.'”
[ToS] states conversations 'is kept until it is no longer needed for operating the service or until deletion is requested' but does not specify concrete timeframes. Deletion requests are permitted but implementation timing is vague ('until deletion is requested' without SLA). Diagnostic logs retention period is mentioned as 'shorter' but not defined, creating moderate uncertainty about data lifecycle.
“[ToS] 'Conversation and archive history is retained for context, search, and operational memory, and is kept until it is no longer needed for operating the service or until deletion is requested. Diagnostic logs are kept for a shorter period.'”
[ToS] explicitly protects user IP: 'You keep any rights you have in what you write.' OpenClaw claims no ownership or license to user-generated content beyond processing for service operation. This unambiguous statement confirms company retains all IP rights to outputs.
“[ToS] 'You keep any rights you have in what you write.'”
[Security] documents security practices including vulnerability disclosure, security audits (OpenAI's Patch the Planet Initiative mentioned Sep 21, 2026), and partnerships with NVIDIA and Tencent for security review, but does not explicitly claim SOC 2, ISO 27001, GDPR compliance certification, or HIPAA. Security practices are documented but formal compliance certifications are not stated.
“[Security] 'OpenClaw Completes Security Audit Through OpenAI's Patch the Planet Initiative' and '[ToS] For privacy requests, security reports, or questions about these terms, contact service@openclaw.org'”
Track and measure your AI tools with Olakai's observability platform.
Explore Olakai