Frequently asked questions

Humasure is a San Francisco startup pioneering independent safety certification for AI products and robots. Products that pass our tests against the 10 Directives will carry the Humasure mark and publish a Trust Label, so people can see at a glance that they stay in charge.

You can always stop it, it was tested, and someone answers for it. Behind that are ten Directives: always stoppable, bounded autonomy, physical safety, honest identity, private by default, secure and supported, no manipulation, fair treatment, accountable, and safe for its whole life.

Not yet. Humasure is in formation and the Directives are a working draft. Any seal or Trust Label you see on this site is a sample. If you see the Humasure mark on a product today, please tell us at info@humasure.org.

It means a person can always stop an AI product or robot, decides what it may do on its own, knows it is dealing with a machine, and can hold a real company to account. It is a principle we share with everyone, not a trademark.

No. Humasure is independent and voluntary. Certification does not mean a product complies with every law; producers stay responsible for the laws of each market they sell in, and where a law is stricter than a Directive, the law applies.

AI software and agents, AI-powered devices, home and service robots, and commercial robots. We certify a specific product, model and software version, not a whole company. We do not certify military or weapons systems.

A stricter level for products marketed to children, or likely to be used by children, older adults or people in care, or used in homes or care settings. It adds requirements such as lower force limits near people, no-record bedrooms and bathrooms, and protections for children and older adults.

Assessors review the producer’s documents, accredited labs test the hardware, and a red team tries to make the AI break its own rules: ignore the stop, act without permission, claim to be human, or manipulate a vulnerable user. The methods for each Directive are listed on the Directives page.

A certificate lasts one year and renews only after surveillance: an annual review, anonymous store purchases, a review of every incident and complaint, and a full reassessment every three years. Material changes, including a new AI model, are reviewed before they reach users.

Depending on how serious it is, the product is certified with conditions, suspended or revoked, and the registry says so publicly. A failure of the stop or of physical safety means immediate suspension. Revocations stay in the registry permanently.

Fees will be published before applications open, set by product category, and will never depend on the result. Robot assessments involve lab testing and cost more than software assessments. Founding Cohort products are certified at cost in exchange for a public case study.

Email info@humasure.org with your product, what it does, who uses it and where it is sold. We are looking for our first ten products in AI companions, eldercare, home robots and AI agents.

Our assessors never advise the producers they certify, to keep the mark independent. Preparation help comes from published, self-serve tools: the Directives, the self-check on our producers page, and a Self-Assessment Toolkit that is in development.

Yes. Products built on general-purpose AI models are in scope. You remain responsible for meeting the Directives, and a new version of the model is treated as a material change if it could affect the stop, autonomy, honesty or manipulation rules.

Yes. The Directives go to founding advisors first, then to a 90-day public comment period. Every comment and our response will be published. Until the comment portal opens, write to info@humasure.org.

A nonprofit Humasure Foundation will own the standard and the mark, with an independent majority on its board, including a consumer advocate. A separate company, Humasure Assurance, Inc., will carry out testing under a license the Foundation can revoke.