AI Terms
These AI Terms (“AI Terms”) are a binding legal agreement between you (“You” or “Customer”) and Emcie Co Ltd. (registration number 516987971) (“Company”), and govern your access to and use of features and functionalities made available as part of the Service, that are based on or powered by artificial intelligence and/or machine learning technologies (such technologies, collectively - “AI”, and such features and functionalities - “AI Features”).
By creating an Account and/or by accessing or using the Service, You accept these AI Terms on behalf of yourself (if You are an individual) or of the company or other organization that You represent, and - as applicable - You agree to be bound by these AI Terms or You represent and warrant that You have full authority to bind the company or organization that You represent to these AI Terms. If You do not agree as aforesaid, or are not authorized as aforesaid, You may not access or use the Service and You must refrain from creating an Account.
Capitalized terms used but not defined in these AI Terms have the meanings set forth in the agreement between Customer and Company governing Customer’s use of the Service, which includes the Terms of Service and documents incorporated into it by reference (the “Agreement”).
These AI Terms are an addendum to, and are incorporated by reference into, Company’s Terms of Service. In the event of a conflict between these AI Terms and the Terms of Service, the order of precedence in Section 19.10 of the Terms of Service applies. Where AI Input or AI Output includes personal data, the Data Processing Addendum shall prevail with respect to the processing of that personal data. For the avoidance of doubt, the dispute resolution mechanism in the Terms of Service shall apply to any dispute arising out of or related to these AI Terms.
1. Definitions:
“AI Input” means data and/or information digitally submitted or made available by Customer, through the Service Interface or through an integration that Company makes available and Customer connects, which is transmitted to an AI Feature, for provision of the Service thereupon. For clarity, input into an AI Feature may consist of both content that belongs to Customer and content that does not belong to Customer (for example, suggested prompts that originate from Company or its service providers). Of all content transmitted as aforesaid to an AI Feature - only the data and/or information originating from Customer shall be considered AI Input. AI Input may include messages, conversation and Session data, source code and repository content from a repository that Customer connects, deployment settings, logs and other activity data, but includes such repository content or other materials only to the extent they are transmitted to an AI Feature through the Service.
“AI Output” means any output generated by an AI Feature based on AI Input and transmitted to Customer or made available to Customer in the Service, including titles, summaries, suggestions (such as suggested build and deployment settings), responses of assistants and generated files. It is clarified, that AI Output specifically excludes: (1) Company’s Intellectual Property in its templates, suggested prompts, proprietary datasets and training materials as such, but not AI Output generated from AI Input merely because it was produced using, or includes parts of, a Company template or suggested prompt; (2) Data provided when establishing the Account; (3) Usage data, meaning data about how the Service is used as described in the Terms of Service, but not the content of AI Input or AI Output; (4) Feedback, whether structured (e.g., "thumbs up/down" signals, ratings) or unstructured (e.g., free-text submissions); and (5) aggregated and de-identified data derived from AI Input or AI Output, but not the AI Input or AI Output themselves or logs that identify Customer or an individual (such data is treated as anonymous only as provided in Section 7.2 of the Terms of Service). The exclusions in (2) to (5) mean only that those items are not AI Output; they do not remove any duty under the Agreement, the Data Processing Addendum or the Privacy Policy that applies to personal data or confidential information contained in them.
2. Customer acknowledges and agrees that the Service is comprised of several AI Features. The use of AI Features for provision of the Service is deemed authorized by Customer upon Customer’s creation of an Account and/or use of the Service, including when Customer or its Authorized Users start an AI Feature. Any such use authorized as aforesaid, constitutes Customer’s agreement to these AI Terms. The AI Features that currently operate, how each is started, the data each uses and the related provider, output, retention and training information are set out in Schedule 1.
3. AI Features authorized by Customer as aforesaid, allow Company to process, either itself or through third parties operating the AI Feature, Customer’s AI Inputs, through the relevant systems, in order to provide the requested functionality. Customer acknowledges that to generate AI Output, the AI Features may analyze Customer’s AI Inputs, which may involve sending the content contained therein to Company’s AI systems or servers or to relevant third party AI providers. The third party AI providers used for each AI Feature are described in Schedule 1 and, where they process personal data, in the Data Processing Addendum. When Customer starts a repository inspection for a Managed Project, the repository content that is transmitted to an AI Feature is AI Input. Charges and terms of any model, tool or key that Customer connects to its own agents are Customer’s, as stated in Section 2.11(f) of the Terms of Service.
4. To the extent an AI Input includes personal data, Company may process it as provided in the Agreement, including the Data Processing Addendum.
5. Customer acknowledges and agrees that use of the AI Features is subject to these AI Terms and the Agreement in whole. In addition to any use restrictions in the foregoing, Customer shall use the AI Features responsibly and in compliance with the following rules:
a) Customer may not use the AI Features in any manner that is unlawful, harmful or violates any applicable law or regulation. The AI Features are to be used only for legitimate purposes (within the bounds of the Agreement) and may not be relied on for any purpose of obtaining advice or instructions to carry out illegal or dangerous activities.
b) Customer may not use the AI Features to infringe upon a right of another. This includes not providing AI Input that violates intellectual property rights, privacy rights or other rights of any person or entity, or that is intended, or would reasonably be expected, to generate AI Output that does so. Customer may not use the AI Features to generate or disseminate content for which Customer does not have the rights or that contains personal data of others without permission. If Customer inputs any data or content obtained from a third party into an AI Feature, Customer must have the necessary rights to use that content. Customer is responsible for any third party content Customer provides as AI Input and for ensuring that using it with the AI Feature (and using any AI Output derived from it) does not violate any law (including, without limitation, privacy and data protection laws) or agreement (such as confidentiality or privacy obligations).
c) Customer may not represent any AI Output as being approved or vetted by Company.
d) Customer is responsible for deciding how AI Outputs are used and for applying checks that are appropriate to the use. This does not require a person to approve each routine response that an agent, which Customer has set up to respond on its own, gives to End Users. Examples of checks that may be appropriate, depending on the use and on Customer’s legal obligations, include: (i) testing and evaluating Customer’s agents and the AI Outputs they rely on before they are launched to End Users; (ii) monitoring and reviewing them on an ongoing basis, for example through the Sessions and logs available in the Service; (iii) providing a way to escalate to a person where the use calls for it; and (iv) stronger checks, which may include human review before an AI Output is acted on or shared, where the use is high-stakes, for example where it supports decisions that significantly affect individuals, or legal, medical, financial or safety matters. Company permits high-stakes uses of the AI Features. Customer is responsible for the safeguards, oversight and requirements that apply to its deployment, according to its use case and its legal obligations, and these AI Terms do not require a person to review every high-stakes AI Output before it is acted on or set a mandatory checklist of safeguards. Nothing in these AI Terms permits Customer to bypass a safeguard that the law requires, states that the Service is ready for healthcare or other regulated use, or commits Company to enter into a business associate agreement. Customer decides which tools, systems and accounts its agents may use and what actions they may take in outside systems, must have the authority and permissions needed for those actions, and is responsible for configuring, limiting and overseeing them. Customer is responsible for providing the AI notices and disclosures required for End Users in the experiences that Customer builds and deploys, including telling End Users, where the law or the circumstances require it, that they are interacting with an AI system, and for the other transparency, oversight and escalation measures that apply to its deployment. Company does not state that using the Service makes a deployment compliant, and these AI Terms do not require a person to approve each external action. Company is responsible for the notices required for its own platform features, including Coparlant and Lighthouse. This allocation does not transfer away any duty that the law places directly on Company or on Customer, and Company does not state that every deployed interface already shows every required notice. Customer remains solely responsible for any consequences of using AI Outputs.
e) Customer may not use any AI Feature to develop a hosted platform or service that competes with Company’s proprietary platform, and may not reverse engineer or extract any component of an AI Feature or its models and algorithms (such as using prompts to discover training data or model weights). Nothing in this paragraph limits Customer’s right to build, change and operate its own agents and products using the Service or Parlant, or any right granted under an open-source license, including the Apache License 2.0 under which the Parlant engine is licensed.
f) Customer may not attempt to manipulate or misuse an AI Feature beyond its intended functionality. For example, Customer will not attempt to trick or prompt the AI Feature into violating these AI Terms (e.g., no “jailbreaking” the AI with malicious prompts). Customer may not introduce any viruses, worms, malware or harmful code into the AI Feature.
6. Company reserves the right to monitor usage of the AI Features for compliance with these AI Terms, and may suspend, limit or terminate Customer’s access to the AI Features if Company believes Customer is in violation of any of them. Company may also set usage limits for AI Features (such as hourly or monthly limits), including for assistive features such as Coparlant, and may refuse further assistance once such limits are reached, as set out in the Documentation. The features available on the Plan that Customer has selected, including managed hosting and builds and platform-provided AI features where the Plan offers them, are included within the limits published for that Plan, and Company does not apply automatic overage charges. Not every feature is available on every Plan. Charges for the model-provider accounts and keys that Customer connects to its own agents are billed separately by those providers, and model spend reporting in the Service is not an invoice from Company.
7. Customer agrees that Company may access and use AI Input and AI Output to operate, evaluate and support the AI Features, comply with applicable law, enforce Company’s terms and policies and keep the Service safe and secure. Use of AI Input or AI Output to improve the AI Features or to train models is not agreed by this paragraph; it depends on paragraphs (d) and (e) below and, for personal data, on the Data Processing Addendum. A Plan’s capture-privacy setting, the former workspace training prompt and a choice about analytics are not permission to use AI Input or AI Output to improve the AI Features or to train models. The following paragraphs explain these uses separately:
(a) Answering requests and providing the AI Features. Company processes AI Input to generate AI Output and to provide the Service, including titles and summaries of Sessions, natural-language search of Sessions, responses of the Coparlant assistant and the inspection of connected repositories. For this purpose Company may send relevant AI Input to third-party AI providers engaged as subprocessors.
(b) Support, security and legal compliance. Company may access and use AI Input and AI Output to provide support, to keep the Service safe and secure, to comply with applicable law and to enforce Company’s terms and policies.
(c) Usage statistics. Company may collect, aggregate and analyze data about usage of the AI Features as provided in Section 7.2 of the Terms of Service. This does not extend to the content of AI Input or AI Output.
(d) Customer-specific improvement. Company’s policy is not to train or tune AI models on AI Input, AI Output or other Customer Data, including training or tuning intended only for Customer, without Customer’s explicit permission and opt-in. That permission is not inferred from creating an Account, general use of the Service, a choice about analytics, a Plan choice or any general license in these AI Terms or the Agreement. Ordinary processing to provide an AI Feature, and any internal observation use that Company separately discloses, are not by themselves authorization to train a model.
(e) Shared model training. Company’s policy is not to train or fine-tune AI models that serve other customers, or general-purpose models, on AI Input, AI Output or other Customer Data without Customer’s explicit permission and opt-in. Analytics consent, a Plan’s capture setting, the former workspace training prompt and the license to Company in these AI Terms are not that permission.
(f) Improvement by AI providers. Whether a third-party AI provider may use data it receives for its own purposes, including improving its own models, depends on the terms that apply to the product, account and contract used for that provider. This applies to providers that Company engages; for a provider whose account, key or tool Customer connects, the provider’s terms are between Customer and that provider (Section 2.11(f) of the Terms of Service). A standard Plan does not promise that AI Input is processed in a particular country or region; only a separate signed Enterprise agreement may do so.
8. As between Customer and Company, Customer retains ownership of its AI Inputs, and - to the extent permitted by law - Customer will own the AI Outputs generated in response to its AI Inputs. Company’s rights in AI Inputs and AI Outputs are limited to the license in the following paragraph. Company hereby assigns to Customer any rights it may have in the AI Outputs (to the extent they exist under applicable law). Notwithstanding the foregoing, Customer acknowledges that AI Outputs may include or be derived from underlying content or patterns from AI or machine learning training data, which could be publicly available information or third party content. Company does not guarantee that every AI Output is entirely unique, and Customer acknowledges that AI Outputs provided to it may be similar or identical to AI Outputs independently provided to others.
9. Customer grants Company, in respect of AI Inputs and AI Outputs, the same license that Section 9.4 of the Terms of Service grants in respect of Customer Data, as if AI Inputs and AI Outputs were Customer Data. That license is limited, worldwide, royalty-free, non-exclusive, non-sublicensable and non-transferable (except to a permitted assignee of Company), and Company may use AI Inputs and AI Outputs under it only for the purposes in Section 9.4 of the Terms of Service, as described in Section 7 of these AI Terms. Company may allow Third Party Providers, including AI Service Providers engaged as subprocessors, to process AI Inputs on its behalf only as provided in Section 9.7 of the Terms of Service. The license does not allow Company to publish, publicly display or distribute AI Inputs or AI Outputs, and showing them to Customer and its Authorized Users in the Service is not publication. Sections 9.7 and 9.8 of the Terms of Service apply to AI Inputs and AI Outputs. If this paragraph and Section 9.4 of the Terms of Service differ as to the scope of Company’s rights in AI Inputs or AI Outputs, the narrower scope applies. The license does not authorize the use of AI Inputs or AI Outputs to train or fine-tune any model; that requires Customer’s explicit permission and opt-in as provided in Section 7(d) and (e).
10. All rights, title and interest in and to the AI Features remain with Company or its licensors, as applicable. These AI Terms do not grant Customer any rights to Company’s Intellectual Property or the Intellectual Property of any third party AI providers, except for Customer’s limited license to use the AI Features as provided herein.
11. Due to the nature of the AI Features, Company does not represent or warrant that: (a) any AI Output does not incorporate or reflect third party content or materials; or (b) any AI Output will not infringe any third party’s Intellectual Property rights. Notwithstanding any indemnification provisions in the Agreement, Company will not indemnify Customer nor any End User for any third party’s Intellectual Property rights claim arising from Customer’s and/or the End User’s use of AI Outputs. Subject to any obligation that cannot be excluded by law, Company does not assume responsibility for AI Outputs and does not give Customer or any End User protection against third party Intellectual Property claims about AI Outputs.
12. The AI Features are provided “as is” and “as available” without any warranties of any kind. To the maximum extent permitted by law, Company disclaims all warranties, express or implied, regarding the AI Features and AI Outputs, including, but not limited to, warranties of accuracy, completeness, reliability, merchantability, fitness for a particular purpose or non-infringement. Company does not warrant that the AI Features will meet Customer’s requirements or that the operation of the AI Features will be uninterrupted or error-free.
13. AI Outputs are generated by AI, which is still improving in accuracy, reliability and safety. When using the AI Features, Customer acknowledges and agrees that: (i) AI Outputs may not always be accurate and may contain material inaccuracies even if they appear accurate because of their level of detail or specificity; (ii) Customer should not rely on an AI Output for a purpose without evaluating whether it is suitable and reliable enough for that purpose, using checks appropriate to the use as described in Section 5(d), which may include human review; and (iii) Company does not promise that any response given by an agent without human approval is accurate, complete or appropriate.
14. Customer acknowledges and agrees that: (i) certain AI Features are provided or licensed by third parties (“AI Service Providers”) and use thereof is subject also to the applicable terms and conditions of the relevant AI Service Provider; and (ii) Company is not liable for AI Features provided or licensed by AI Service Providers or for their outcomes, except that this does not limit Company’s responsibility for its subprocessors under Section 8.4 of the Data Processing Addendum.
15. Customer agrees, at its own cost, to defend Company from and against any Covered AI Claims (as defined below) and to indemnify and hold Company harmless from and against any damages and costs awarded against Company (including reasonable attorneys’ fees) resulting from Covered AI Claims. This indemnity is in addition to any indemnification obligations in the Agreement. This indemnity is not limited by Section 15 of the Terms of Service.
“Covered AI Claims” mean any third party claims arising out of or related to Customer’s use of the AI Features, the AI Inputs or AI Outputs, or Customer’s breach of these AI Terms.
16. These AI Terms may be modified by Company from time to time, by posting an updated version thereof on the Website. Customer is responsible for regularly reviewing the most current version thereof. When modified - Company will update the “Last Updated” date stated on it. By continuing to use the Service after the effective date of any such modification, Customer will be deemed to have accepted such modification. Material modifications are notified as provided in Section 19.7 of the Terms of Service.
Schedule 1 – AI Features
| AI Feature | How it starts | Data used | Provider and keys | Output, retention and training |
|---|---|---|---|---|
| Session titles and summaries | Titles: automatic, from stored Sessions of Live Projects, written by the Ingestor once a Session has been quiet for a delay. Summaries: in the platform mode that writes summaries with titles, written automatically with the title; otherwise generated when a summary is requested for a Session. Not available for simulation Sessions, and a Local Project’s Session is not summarised. | Stored conversation interactions of the Session | A Gemini integration in the Ingestor. | Title and summary shown with the Session. Retention: Plan history and retention limits govern what Customer can see and do not guarantee physical deletion. Company retains full conversations and related logs internally for 24 months after the last activity in the Workspace, for usability improvement and safety purposes, and then deletes them; this content is not described as anonymized or reduced, and access is limited to Company admins and designated support and safety staff (Company staff, not Workspace roles). Company will delete retained content on a valid explicit privacy-deletion request sent separately to [email protected]; in a shared Workspace the request must be authorized by an Owner. Plan expiry and deleting a Project in the dashboard are not such a request. Training: Company’s policy is not to train AI models on this content without explicit permission and opt-in (Section 7(d) and (e)); any provider’s own training terms are addressed in Section 7(f). |
| Coparlant assistant | Started by the user in the Service | Questions entered by the user, and the agent and Session data the assistant is asked to examine, which it reads through authenticated backend tools in the context of the user’s Project | Configured by Company in the deployment. The default is Gemini, and OpenAI is also supported as a configured alternative. | Responses shown to the user, subject to the Plan’s limits. Retention: Plan history and retention limits govern what Customer can see and do not guarantee physical deletion; as for Session titles and summaries above, Company retains full conversations and related logs internally for 24 months after the last activity in the Workspace, and deletes retained content on a valid explicit privacy-deletion request sent separately to [email protected]. Training: Company’s policy is not to train AI models on this content without explicit permission and opt-in (Section 7(d) and (e)); any provider’s own training terms are addressed in Section 7(f). |
| Repository discovery for managed deployments | Started by the user for a connected repository and branch | Repository files at the checked commit; detected build and run commands; ordinary variable names and values and the names of required secrets found there. Platform-saved secret values are not designed to be supplied to the inspection; this is not a guarantee that no secret can reach the AI provider. Files committed to the repository can be read, so a secret committed to a repository may be exposed to the inspection. | Claude, through the Claude Agent SDK in the deployment service, with the model and effort set by deployment configuration (currently Claude Haiku at medium effort). | Detected commands, variables and required secret names, which fill fields the user has not yet edited in the initial setup and are otherwise offered as suggestions that the user accepts or rejects. The user reviews them before deploying. A discovery result remains readable for 30 days after it ends. Training: Company’s policy is not to train AI models on this content without explicit permission and opt-in (Section 7(d) and (e)); any provider’s own training terms are addressed in Section 7(f). |
| Lighthouse documentation assistant (public) | Started by a caller of the public query endpoint or the command-line tool, with or without a dashboard account | The question, relevant context and client and SDK version details supplied by the caller, after pattern-based redaction of secrets (redaction does not mean the text is free of personal data). The assistant reads a versioned snapshot of the documentation through search and read tools. | Google Agent Development Kit. Answers are generated with Anthropic’s Claude Opus through the Anthropic API; Google Vertex AI is used for embeddings, answer processing and evaluation; Google Firestore is Lighthouse’s database (this does not mean the wider platform has stopped using MongoDB Atlas). | The answer, with citations, is returned to the caller. Redacted questions, context, answers, duration, usage and cost estimates are stored for staff review. Retention: records are retained for 24 months after the last activity on the record (the date of the query) and may be read by Company admins and designated support and safety staff. The assistant can be used anonymously. When a caller supplies optional authentication and it is verified, a saved query record may also carry the caller’s user ID and name and the workspace (tenant) ID and name, and the same staff can read them. The company uses this identity information for proactive support and troubleshooting, to attribute queries to a user or workspace, and to remove a person’s records on request. Questions, supplied context and answers are only secret-redacted and free text may still contain personal information, so the records are not described as anonymous. The service returns a request ID for each saved query; a requester who has it can give it to help locate a record, but it is not the only way a record may be located. Company will honor deletion requests sent to [email protected] where the relevant records can be identified and the request verified. Separately from these query records, the released Lighthouse code can rewrite suitable answers into general Parlant questions and answers, remove request-specific details, create embeddings and save the result for reuse, and a link to the source record is kept. This reuse is not model training, and the saved result is not described as anonymous or as impossible to delete. |
| Natural-language search of Sessions | Automatic indexing when a Session’s title or summary changes, and a search query started by a user in the Service | Search text made of the Session’s display title (which can be a title set by a person, a generated title, the source’s title or an excerpt of the first message) plus the summary when present, up to a maximum length; and the query text. | The database service’s automated embedding (MongoDB Atlas) using the Voyage model voyage-4-lite. The application does not call an embedding API itself, which does not mean that no model processes the text. Separate from the Gemini route for titles and summaries. | Matching Sessions with relevance scores, limited to the Workspace and Project. The search text and its embedding are kept in a search index. Retention: search entries are deleted when the underlying Session content is deleted under Section 9.6 of the Terms of Service. Training: Company’s policy is not to train AI models on this content without explicit permission and opt-in (Section 7(d) and (e)); any provider’s own training terms are addressed in Section 7(f). |
| Customer’s own agent | Whenever the agent handles a conversation | As configured by Customer | Not an AI Feature of Company unless stated. The model, provider and keys chosen and configured by Customer; the provider’s terms apply to Customer. A provider chosen and connected by Customer is not Company’s subprocessor merely for that reason. | Customer’s responsibility |
Company admins (Company’s own staff, not Customer roles) are responsible for keeping this Schedule and the public lists of AI Features and providers accurate.
Last Updated: November 1, 2026