Use of AI Terms

The AI Features help your team draft, summarize and edit. As of the date at the top of this page they include:

• Workflow drafting. A chat that turns a plain-language description of how your business works into a draft workflow for you to review.
• Message drafting. Suggested replies and summaries for conversations in your inbox.
• Image editing. Edits and renders of photos you upload, such as before-and-after images of a vehicle.

AI Features are designed to assist and augment your work, not to replace human judgment and decision-making. Every output is a draft for a person on your team to check.

All AI features run through a single inference provider, OpenRouter, which passes each request to one of a number of third-party model providers. We do not publish which model answers a given request, because the choice changes as models are added, retired and improved, and we would rather change it than have this page go out of date.

When you use an AI Feature, the content you submit ("Input"), such as a prompt, a workflow brief, a conversation, or an image, is sent through OpenRouter to a model provider to generate a response ("Output").

Two rules govern where a request may go. We route only to model providers whose terms prohibit training on the content sent to them. And we route only to providers that do not retain or log that content in the ordinary course, so an Input and its Output are used to answer the request and are not kept afterwards.

Logging is off unless we deliberately turn it on. We may switch on logging for a specific, time-limited purpose, such as diagnosing a fault you have reported or building and tuning a workflow with you. When we do, it captures only what that task needs, it is switched off when the task is finished, and anything captured is held under the same terms as the rest of your data and is never used for training.

The current list of model providers we are willing to route to is available on request, and the sub-processor annex to your Data Processing Addendum records the arrangement for your organization.

We engage AI providers under business terms that prohibit them from using your Input or Output to train or improve their general models. We may change or add providers; if we do, we will update this page and, where the change affects Customer Data, the Subprocessor List in our Data Processing Addendum.

Ownership. As between you and flow, you own your Input. To the extent permitted by law, you also own the Output generated from it, subject to the rights of any third party whose material was included in your Input. flow retains ownership of the AI Features themselves, including the prompts, code and configuration we use to operate them.

No training. We do not use your Input or Output, whether identifiable or de-identified, to train, fine-tune or improve any AI model, ours or a third party's. Where your Input includes personal data about your customers or staff, the commitments in Section 10 of our Data Processing Addendum apply.

Similar output. Because models generate responses from patterns, other customers may receive Output that is the same as or similar to yours. Output is not exclusive to you and we make no promise that it is unique.

Copyright. Output generated by AI may not be eligible for copyright protection in all jurisdictions. If that matters for how you plan to use an Output, take advice before relying on it.

AI Features can produce Output that is wrong, incomplete, out of date, or that looks confident and is not. You are responsible for reviewing Output before you rely on it, send it to a customer, or act on it. Do not use Output as a substitute for professional advice in areas such as law, tax, medicine, or safety.

Where an AI Feature drafts a message to one of your customers, the message is not sent until a person on your team chooses to send it, unless you have configured a workflow to send automatically. If you configure automatic sending, you are responsible for what is sent.

In addition to the restrictions in the Terms of Use, you may not use the AI Features to:

• Generate content that is illegal, harmful, harassing, discriminatory, or that violates the rights of others, including their privacy and intellectual property rights.
• Create misleading content, impersonate a person or organization, or produce images of real people without their consent.
• Send unsolicited messages, or messages that breach the terms of the messaging platform you are using.
• Make decisions about a person that have legal or similarly significant effects on them, such as decisions about employment, credit, insurance, housing or access to essential services, without meaningful human review.
• Submit Restricted Data as defined in our Data Processing Addendum, such as government ID numbers, health records, payment card numbers, or data about children.
• Attempt to extract, reverse engineer, or replicate the models, or to bypass safety measures or rate limits.
• Use the AI Features in a way that violates the acceptable use policy of the model provider, including Google's Generative AI Prohibited Use Policy.

We may suspend access to the AI Features, or to the Service, if we believe these rules are being broken.

The AI Features do not make decisions that produce legal or similarly significant effects on any person. They draft, suggest and summarize for review by your team. If we ever introduce an AI Feature that does make such decisions, we will tell you before you can enable it and explain how it works, as described in our Data Processing Addendum.

We are continuously developing the AI Features. We may add, change, or discontinue an AI Feature, or change the provider behind it, with reasonable notice where practical. New AI Features are subject to these AI Terms unless we say otherwise when we release them.

The disclaimers and limitation of liability in the Terms of Use apply to the AI Features and to any Output. Nothing in these AI Terms increases our liability beyond what the Terms of Use provide.

If you have questions about the AI Features or these AI Terms, email us at team@takeflow.io.