AI Terms
Last changed 20 August 2026
In brief
Telos Brain uses AI models to read your organisation's information and generate answers. You supply the models. You bring your own accounts and API keys with the model providers you want to use, and you pay them directly. That means the model provider's data practices — including whether it trains on what you send it — are governed by your agreement with that provider, not ours. We tell you what we can and cannot commit to below.
Answers can be wrong, and sometimes confidently wrong. You need to check anything that matters before you act on it, and you must not let the Brain make decisions about people on its own. We do not use your information to improve anything we offer to other customers.
1. Scope
1.1 These AI Terms apply to your use of any feature of the Service that uses artificial intelligence or machine learning, and to all AI Outputs, in both a Cloud Deployment and a Self-Hosted Deployment.
1.1A These AI Terms are published by Telos Limited for the Telos Group. They are incorporated into, and form part of, your Terms of Use with Telos NZ Limited. References to "we", "us" and "our" mean Telos NZ Limited as your contracting party, and Telos NZ Limited is accountable to you for compliance with these AI Terms.
1.2 Capitalised terms have the meaning given in the Terms of Use. In these AI Terms, "Terms of Use" means the Telos Brain Terms of Use.
1.3 Clause 9 of the Terms of Use sets out the core position, including the fact that you supply your own models and Model Credentials. These AI Terms expand on it. Where there is any inconsistency about AI Outputs, Model Providers or model behaviour, these AI Terms prevail.
2. How the Service uses AI
2.1 The Service uses AI models to interpret, organise, index, embed, summarise, retrieve and generate content from Your Content and from the instructions you give it, and to execute Skills.
2.2 The Service does not include an AI model. All model operations — including large language model inference and the generation of embeddings — are performed by Model Providers, using Model Credentials that you obtain and configure. To generate an AI Output, a Brain transmits parts of Your Content, including prompts, retrieved context and conversation history, to the Model Provider you have selected.
2.3 Where that transmission occurs depends on your Deployment:
- in a Cloud Deployment, the call is made from our infrastructure to the Model Provider, using the Model Credentials you have configured; and
- in a Self-Hosted Deployment, the call is made from your environment directly to the Model Provider. It does not pass through our systems, and we do not receive Your Content, your prompts, your AI Outputs or your Model Credentials.
2.4 The Documentation lists the Model Providers the Service currently supports. We may add or remove connector support at any time. Adding or removing connector support does not change who you contract with: you contract directly with each Model Provider you choose to use.
2.5 Our sub-processors — the third parties who process Your Content on our behalf in a Cloud Deployment, including other members of the Telos Group — are published on the Trust Centre at https://www.telosbrain.com/legal/data-processing. Model Providers you connect using your own Model Credentials are not our sub-processors (see clause 5.6 of the Data Processing Addendum). We will give you notice of changes to our sub-processors in accordance with clause 5.3 of the Data Processing Addendum.
2.6 Models change. Model Providers update, retrain, deprecate and withdraw models, and you may change which model a Brain uses for a given task at any time. This means the Service may produce different AI Outputs for the same input at different times, and behaviour you have come to rely on may change. We do not warrant consistency of AI Outputs over time, and we are not responsible for a change made by a Model Provider.
2.7 Where we make an AI feature available on a beta, preview or evaluation basis, clause 4.8 of the Terms of Use applies to it.
3. Accuracy and reliance
3.1 AI Outputs are generated by statistical models. They may be inaccurate, incomplete, outdated, internally inconsistent, or entirely fabricated while appearing authoritative. This is an inherent characteristic of the technology and not a defect in the Service.
3.2 We do not warrant the accuracy, completeness, currency, reliability or fitness for purpose of any AI Output.
3.3 AI Outputs are not professional advice of any kind, including legal, financial, accounting, tax, investment, medical, clinical, engineering, safety or employment advice, and must not be relied on as a substitute for advice from a qualified professional.
3.4 An AI Output may cite, quote or attribute a source. Citations may be inaccurate or invented. You must verify a citation before relying on it.
3.5 The Service may generate AI Outputs from Your Content that is itself inaccurate, out of date or contradictory. The quality of AI Outputs depends on the quality of Your Content, and you are responsible for the accuracy of Your Content under clause 6.3 of the Terms of Use.
3.6 The quality, safety behaviour and cost of AI Outputs also depend on which Model Provider and model you select and how you configure it. Those are your choices, and we are not responsible for them.
4. Your obligations when using AI features
4.1 You must review each AI Output before relying on it or acting on it, to a standard appropriate to the consequences of the decision it informs. This obligation is not satisfied by reviewing AI Outputs on a sample basis where an individual output drives a consequential decision.
4.2 You must not use the Service, and must ensure your Permitted Users do not use the Service:
- as the sole or determinative basis for any decision that has a legal, financial, health, safety, employment, educational or housing consequence for any person;
- to make, or to substantially automate, any decision about an individual without meaningful human review by a person with authority and capacity to reach a different conclusion;
- to profile, score, rank or assess individuals, other than where you have your own lawful basis to do so and a human decision-maker remains responsible;
- to generate content that is presented to any person as having been produced by a human, where that person would reasonably expect to be told it was AI-generated;
- in any setting where a failure or inaccuracy could result in death, personal injury, or severe environmental or property damage;
- for any practice prohibited by clause 3.2 of the Acceptable Use Policy; or
- for any purpose otherwise prohibited by the Acceptable Use Policy.
4.3 You must ensure that your Permitted Users are made aware that the Service uses AI and that AI Outputs require review.
4.4 Where you make AI Outputs available to your own customers or to the public, you are responsible for any disclosure obligations that apply to you, including obligations to disclose that content is AI-generated or that an interaction is automated.
4.5 You are responsible for configuring the scope of what a Brain can access and do, including which data sources it retrieves from and which actions any agent or Skill you configure may take. Where a Skill or agent can take an action with an effect outside the Brain — including sending a message, writing to a connected system, or incurring a Model Provider Charge — you are responsible for that action.
5. Training, learning and improvement
5.1 We will not use Your Content to train, fine-tune, evaluate or otherwise improve any model, Skill or capability that we make available to any other customer, unless you have given us separate express written consent.
5.2 The Service will build and refine memory, indexes, embeddings, structure and Skills within your Brain from Your Content. That material forms part of Your Content under clause 7.1 of the Terms of Use and is used only to provide the Service to you.
5.3 What we cannot commit to. Because you contract with each Model Provider directly, using your own Model Credentials, we have no contract with that Model Provider in respect of your use and cannot impose obligations on it. We therefore give no commitment, and can give none, about whether a Model Provider retains, discloses or uses your prompts, retrieved context or AI Outputs, including whether it uses them to train or improve its own models. That is governed by your agreement with that Model Provider and by the settings on your account with it.
5.4 What you should do. Before you configure a Model Provider, review its data usage terms and the data controls available on your account tier. Most major providers offer commercial or enterprise terms under which they do not train on data submitted through the API, and some offer zero-retention options. Whether those apply to you depends on the account and plan you hold. Where the Documentation records our understanding of a supported Model Provider's default position, that information is provided for convenience only, may change without notice, and you must verify it yourself.
5.5 In a Self-Hosted Deployment, we never receive Your Content, your prompts or your AI Outputs at all, and clauses 5.1 and 5.2 operate accordingly.
5.6 We may use aggregated statistical information about the operation of AI features, in accordance with clause 7.5 of the Terms of Use, provided it does not identify you, any individual, or any of Your Content. In a Self-Hosted Deployment, that information is derived only from Metering Data.
5.7 In a Cloud Deployment, we may retain prompts, retrieved context and AI Outputs for a reasonable period for the purposes of providing the Service, diagnosing faults, investigating abuse and meeting our legal obligations. Retention periods are set out in clause 7.1 of the Privacy Policy. In a Self-Hosted Deployment, retention of that material is entirely within your environment and under your control.
6. AI Outputs and intellectual property
6.1 As between you and us, AI Outputs generated for you form part of Your Content and you own them to the extent they are capable of ownership. Your rights in an AI Output may also be affected by your agreement with the Model Provider that generated it, and you should check its position on output ownership.
6.2 AI Outputs are not necessarily unique. The same or substantially similar AI Output may be generated for another customer from a similar input, and we do not restrict our other customers from using AI Outputs generated for them.
6.3 The legal status of AI-generated material is unsettled in many jurisdictions. We make no representation that any AI Output attracts copyright or any other Intellectual Property Right, that you can register or enforce rights in it, or that its use will not infringe the rights of a third party.
6.4 The indemnity in clause 17.3 of the Terms of Use does not apply to AI Outputs, or to any model or output of a Model Provider. Some Model Providers offer their own indemnity for outputs generated through their models; whether one is available to you is a matter between you and that provider. You are responsible for reviewing AI Outputs for third-party rights before using them, particularly where an AI Output reproduces or closely resembles existing material.
7. Confidentiality and security of AI processing
7.1 Your Content transmitted to a Model Provider remains your Confidential Information as between you and us, and is subject to clause 12 of the Terms of Use. Its treatment by the Model Provider is governed by your agreement with that Model Provider.
7.2 Model Credentials are your Confidential Information. Clause 9.14 of the Terms of Use sets out how we handle Model Credentials you configure into a Cloud Deployment. In a Self-Hosted Deployment we never receive them and you are responsible for their security. Do not send Model Credentials to us in a support request.
7.3 Prompt injection. AI models are susceptible to prompt injection, in which instructions embedded in content cause a model to behave in a way you did not intend. Where you connect the Service to a Third-Party Service, ingest content from an untrusted source, or allow a Skill or agent to take actions, you accept that risk and are responsible for limiting what the Brain can access and do. Prompt injection can also cause unintended Model Provider Charges, for which clause 9.11 of the Terms of Use applies.
7.4 You must not attempt to cause the Service to disclose our system prompts or instructions, another customer's content, or any part of the Platform, and must not attempt to circumvent any safety measure. Clause 6.2 of the Terms of Use and the Acceptable Use Policy apply.
8. Regulatory roles and AI-specific law
8.1 A number of jurisdictions now regulate AI systems and allocate obligations between the party that supplies a system and the party that puts it into use. This clause allocates those roles between us. It does not change either party's obligations to a regulator.
8.2 As between us, and to the extent an AI-specific law applies:
- we supply the AI system. Where the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) applies, we act as the provider of the Telos Brain AI system; and
- you put it into use under your own authority, determine its purpose, decide what data it processes, select the models it calls, and configure the Skills and agents it runs. Where the EU AI Act applies, you act as the deployer.
8.3 We are responsible for provider-side obligations in respect of the Software and the Cloud Service as we deliver them, including designing the Service so that a natural person interacting directly with it is informed that they are interacting with an AI system, where that obligation applies and is not obvious from the context.
8.4 You are responsible for deployer-side obligations arising from how you use the Service, including:
- assigning human oversight to persons with the necessary competence, training and authority;
- ensuring input data is relevant and sufficiently representative for your intended purpose;
- monitoring operation and keeping any logs your obligations require;
- informing workers and their representatives before putting an AI system into use in the workplace;
- informing individuals where you use the Service to generate or manipulate content that constitutes a deep fake, to produce text published to inform the public on matters of public interest, or in connection with emotion recognition or biometric categorisation; and
- carrying out any fundamental rights or data protection impact assessment required of you.
8.5 You may become a provider. If you place the Service on the market under your own name or trade mark, make a substantial modification to it, or put it to a use classified as high-risk, you may yourself become the provider of an AI system for regulatory purposes and assume the corresponding obligations. If you intend to do any of those things, you must tell us in advance. We give no warranty that the Service is suitable for a high-risk use.
8.6 You must not use the Service for any practice prohibited by Article 5 of the EU AI Act or by an equivalent prohibition in any jurisdiction that applies to you. Clause 3.2 of the Acceptable Use Policy lists those practices.
8.7 You are responsible for your own compliance with any other AI-specific law that applies to you, including laws governing automated decision-making, consequential decisions, algorithmic discrimination and AI disclosure. We will provide information reasonably available to us to assist you, at your cost where the request is substantial or repeated.
9. Changes to these AI Terms
9.1 We may change these AI Terms in accordance with clause 2 of the Terms of Use. Superseded versions are archived at https://trust.telosready.com.