AI Use Policy

    Effective: 2026-04-20

    This AI Use Policy describes how Unshift LLC ("Unshift", "we", "us") uses artificial-intelligence models in the Services, who owns the output, and the limits of what you can expect from AI-generated content. It is part of the Terms of Service and supplements the Privacy Policy and the Acceptable Use Policy.

    If anything in this policy conflicts with a mandatory consumer-protection right in your country, the mandatory right prevails.

    1. What AI does in Unshift

    The Unshift Studio uses AI models from OpenAI and Anthropic to:

    • Generate website copy (headlines, body text, microcopy) from prompts you submit.
    • Generate or refine HTML/JSX/CSS for pages and sections.
    • Propose design choices (color, typography, layout) based on your description.
    • Power conversational editing ("make this section more formal", "add a pricing table").
    • Run internal agent workflows that plan, write, and iterate on changes.

    When you trigger an AI action, your prompt and relevant context from your project are transmitted to the AI provider, which returns generated content ("AI Output"). We then display, store, and/or deploy that Output as part of your site.

    2. Your inputs

    2.1 Prompts and context

    "Your Inputs" are the prompts, instructions, reference materials, brand assets, and project content you submit to the AI features. You represent that you have the right to submit those Inputs and that doing so does not infringe anyone else's rights.

    2.2 What we do with Your Inputs

    • We transmit them to the selected AI provider over TLS.
    • We store them in your project so that you can review, reuse, and edit.
    • We use them to operate the Service on your behalf, as described in the Privacy Policy.

    2.3 What we do NOT do

    • We do not use Your Inputs to train any machine-learning model, whether our own or a third-party model.
    • We do not sell Your Inputs or share them with third parties other than the AI providers you are actively using.
    • We do not publish Your Inputs. Only your published site content is public, and only content you choose to publish.

    2.4 AI provider training

    Our API integrations with OpenAI and Anthropic are governed by their enterprise/API terms, which prohibit those providers from using API-submitted data to train their foundation models. They may retain inputs and outputs for a limited period (typically up to 30 days) for trust-and-safety review and abuse detection, after which the data is deleted. Links:

    If you need stricter handling (for example, zero-retention), contact sales@unshift.ai to discuss a Business-plan arrangement with appropriate provider enterprise agreements.

    3. AI Output disclaimer

    3.1 AI Output may be wrong

    Large language models generate plausible-sounding text and code. They make mistakes. They can invent facts, misstate figures, produce subtly broken code, misquote sources, misspell proper nouns, apply the wrong regulation to the wrong country, and contradict themselves from one request to the next. You are responsible for reviewing all AI Output before relying on it, publishing it, showing it to a customer, submitting it to a regulator, or making a business decision based on it.

    Do not treat AI Output as:

    • Legal advice, financial advice, medical advice, tax advice, or professional advice of any kind.
    • Factually verified information. It is a starting draft; verify with primary sources.
    • Code that has been reviewed for security, privacy, accessibility, or licensing compliance.
    • A substitute for a human editor.

    3.2 No representations or warranties

    To the maximum extent permitted by law, we make no representation or warranty that AI Output is accurate, complete, current, fit for any particular purpose, or non-infringing. AI features are provided on an "as is" and "as available" basis.

    3.3 Your responsibility to review

    Before publishing AI Output on a live site, using it to contact customers, or embedding it in a product that people will rely on, you must review it. Publishing unreviewed AI content that harms a third party (false claims, defamatory statements, leaking of unrelated customers' data accidentally pulled from the model's training, etc.) is your responsibility, not ours. The Terms of Service §13 indemnification applies.

    4. Ownership of AI Output

    4.1 As between you and Unshift

    As between you and us, you own the AI Output generated for your project, to the extent ownership is legally possible. We grant you the rights needed to use, modify, publish, and commercialize the AI Output as part of your Service usage, subject to the Terms of Service and this Policy.

    4.2 Legal reality of AI ownership

    Copyright law for AI-generated content is unsettled. In the United States, the Copyright Office's current position (as of 2024 guidance) is that purely machine-generated output is not copyrightable by the prompting human. Where there is substantial human creative contribution (selection, arrangement, editing, combination with your own authored material), the combined work may be copyrightable to the human author. Similar uncertainty exists in the EU and UK.

    Practical implication: you should not assume that AI-generated text or code is automatically protectable under copyright law in every jurisdiction. If exclusivity matters for your project, add human authorship, edit substantially, or seek legal advice.

    4.3 Not exclusive to you

    Because AI models generate probabilistically, other users may receive similar or identical output for similar or identical prompts. We cannot and do not guarantee that AI Output is unique to you. You should not publish AI Output that must be unique (for example, a registered trademark, unique brand name, or protected creative work) without human editing sufficient to make it distinctive.

    4.4 Third-party rights in AI Output

    AI models are trained on large corpora that can include third-party content. Output may, in rare cases, substantially reproduce protected expression from training data. You are responsible for not publishing AI Output that infringes a third party's rights. If you receive a DMCA notice or similar complaint, treat it as you would any other copyright claim (see Terms §8).

    5. Responsible use of AI features

    You agree not to use AI features to:

    1. Generate content that violates the Acceptable Use Policy, including CSAM, non-consensual sexual imagery, malware, phishing content, or content that incites violence.
    2. Impersonate a real person without their consent, or generate content attributed to a real public figure that the person did not say or do.
    3. Generate content that facilitates fraud, including fake reviews, fake testimonials, misleading medical or financial claims, or misleading product claims.
    4. Extract system prompts, jailbreak the models, bypass safety filters, or exfiltrate other users' data.
    5. Use AI Output to train or evaluate a competing product, or to otherwise build a model or service that competes with Unshift.
    6. Submit prompts in languages, formats, or payloads designed to consume disproportionate credits or compute.

    Violations may result in the enforcement actions described in the AUP §6, including suspension and termination.

    6. Credits and AI usage

    Each AI action consumes credits from your balance, priced per published site and per credit pack. Credits are consumed at the moment of generation, whether or not the output is useful to you. Iterating with prompts is the normal way to use the Service; additional iterations consume additional credits. See the Refund Policy §3 for when credits are and are not refundable.

    We reserve the right to rate-limit or temporarily block AI features if we detect abusive usage patterns (abnormal volume, prompt injection, or other misuse), and to contact you before the issue escalates.

    7. Transparency

    7.1 Which provider is used

    The Studio may route a given generation to OpenAI, Anthropic, or a specific model variant based on task, quality, and availability. We may change providers and routing without notice, provided the overall service quality is maintained. Enterprise customers with specific provider requirements should contact sales@unshift.ai.

    7.2 Disclosures on your sites

    If you publish AI-generated content on a consumer-facing site in a jurisdiction that requires AI-content disclosure (for example, specific US state laws or the EU AI Act's transparency requirements for certain kinds of content), you are responsible for adding the required disclosure. Unshift does not automatically label generated content.

    7.3 Automated decision-making

    We do not use AI to make decisions about you that have legal or similarly significant effects. AI features in the Studio are interactive and always subject to your review and publication choices.

    8. Security of prompts

    AI features transmit Your Inputs to third-party model providers. If your Inputs contain secrets (API keys, passwords, customer personal data beyond what is needed), those secrets are transmitted to the provider. Do not include secrets in prompts. Put secrets in environment variables and reference them from code, or use a dedicated secrets manager.

    If you accidentally submit a secret in a prompt, rotate the secret immediately and contact security@unshift.ai so we can help you assess the exposure.

    9. Changes to this policy

    We update this AI Use Policy as the models, providers, and regulatory environment evolve. Material changes are announced at least 14 days before taking effect. The current version always lives at https://unshift.ai/ai-policy.

    10. Contact