Steadrun AI Disclosure
Last updated: 2026-09-08.
中文摘要：Steadrun 用 AI 和规则自动化替您经营您自己的公司。它是您的电子代理：公司归您，责任归您，它在您授权的范围内以您的名义行事。它会出错。自动运行不等于您退出——任何时刻您都可以介入任何一步，引擎继续跑其余部分。凡是花钱或对外的动作默认要您点一次；每个完成的任务都有证据；平台要求的 AI 内容标识我们默认打开。它不是律师、会计或医生。
Your electronic agent. The Services plan and execute work for a business that is yours: you own it and everything created for it; the Services act in your name within the authorization tiers, budgets and credentials you set; what they do within that authorization is attributed to you. Steadrun owns only the engine.
Automation with intervention. Automated operation and your participation are not alternatives. While the Services run, you can at any moment inspect every role, chain step and piece of evidence; approve, cancel or defer any task; change any tier; switch a chain to step-by-step approval; instruct the decision agent; pause, freeze or shut the company down; revoke credentials; export everything. Interventions take effect immediately for new actions and are recorded in an append-only ledger; the Services keep running whatever you did not touch.
What is automated. Planning of daily tasks, research, content production, page generation, inbox triage and replies, proposals, reconciliation, demand signals, stage-gate checks and reports are produced by rule-based automation and large language models.
What is never automated. Refunds, payouts, transfers, contracts, account opening, payment credentials, first OAuth consent, and regulated advice. Actions that spend money or reach the public are held for your approval unless you have explicitly set them to run automatically.
Accuracy. AI outputs can be wrong, outdated or inappropriate while appearing confident. Review before relying on them; every completed task links to evidence so you can check what actually happened.
Disclosure to your customers. Replies sent from your inbox are signed with your company's name and can state that they were prepared with automated assistance. Media the Services publish carries the platform's synthetic-content flag by default. Notices to your own customers about automated systems are yours to publish; we supply templates.
Models. Steadrun currently uses Anthropic Claude models through the Claude API / Claude Code, in configurations that do not train on your data.
Questions: ai@steadrun.com (until the domain is live: steadrun@aurora-studio.net).