AI Terms

• AI is off by default. AI comes in two packages: Voyager 1 (free) and Voyager 2 (credit-based). Both are disabled by default. Each is turned on by an administrator; Voyager 2 is enabled in the Automation Hub with per-role controls and spending caps. See "The Voyager packages" below.
• Your data is never used to train AI models and is never sent to an outside AI service. AI runs inside Coevera's own AWS environment via Amazon Bedrock. This is reflected on our Security page.
• AI inference currently runs in the United States (AWS US East, N. Virginia) for all customers. Your CRM data at rest stays in the AWS region you select; when AI Features are used, the AI request is processed in the US. EU-resident AI inference is planned but not yet available (see §8).
• You own the AI outputs. They are your business data, like anything else in your CRM.
• AI can be wrong. Outputs must be reviewed by a person before you rely on them.
• Acceptable-use rules apply: ours (§2) and our model providers' (incorporated by reference in §2). They are written to work in any jurisdiction and to satisfy AI-specific laws including the EU AI Act and US state AI laws.

Your responsibilities come first

Sections 2 to 5 set out what you must do when using AI Features: acceptable use, reviewing AI outputs, disclosing AI to your own audience, and your own obligations under AI law. They lead this Addendum because they matter most for compliant use. Sections 6 to 10 explain how the AI Features work and our operational terms.

This AI Features Addendum ("Addendum") supplements the agreement under which your organisation ("you", "Customer") uses the Coevera Services (the "Agreement"), as set out in our Terms and Conditions. It governs your use of the AI-assisted capabilities of the Coevera product, delivered as Voyager 1 and Voyager 2 (together, the "AI Features"). It is accepted on your organisation's behalf by an administrator with authority to do so, when AI Features are first enabled or used for your account. It binds you and the people who use your account ("your users"). The "At a glance" summary and the FAQ on this page are explanatory aids; if they conflict with this Addendum, the Addendum prevails.

You will not use the AI Features, and will ensure your users do not use them, to:

1. violate applicable law, or infringe third-party rights (including intellectual-property and privacy rights);
2. cause harm to minors, generate child sexual abuse material, or sexualise minors in any form;
3. develop weapons, malicious code, or attacks on systems, networks or critical infrastructure;
4. deceive people, including presenting AI output as human-authored where this misleads, impersonating real persons, or generating disinformation or fraudulent content;
5. make decisions with legal or similarly significant effect on a person (for example employment, credit, insurance, legal or health decisions) without qualified human review of each such decision and disclosure to the person concerned that AI was involved;
6. operate any practice prohibited by AI laws applicable to you (manipulative, exploitative, or unlawfully discriminatory AI practices, or practices prohibited by Article 5 of the EU AI Act or analogous provisions of other jurisdictions' AI laws), or deploy a high-risk or consequential-decision AI system without completing the obligations that apply to you;
7. extract model inputs/outputs at scale to train competing AI models ("model scraping" / distillation), probe or circumvent safety guardrails, or attempt to extract the underlying models.

Provider policies (incorporated by reference). Your use of AI Features is additionally subject to the acceptable-use policies of the model providers and gateways we use, as updated by them from time to time: the Anthropic Usage Policy, the AWS Responsible AI Policy, and the model terms collected on the AWS Bedrock third-party model terms page. Where a material upstream change affects what you may do with the AI Features, we will reflect it on this page and notify you per §9. If this Addendum is stricter than a provider policy, this Addendum applies.

As between you and Coevera, AI-generated outputs are yours and become part of your CRM data. You must make this known to your users. Discrete generative artefacts (generated images, AI-drafted text, AI summaries) are marked as AI-generated in the product interface. AI-assisted field values are your responsibility downstream. Values that AI Features write into structured CRM fields (for example AI Smart Field) become part of your CRM records and are handled like any other field value you maintain; they are not separately marked as AI-generated in downstream depictions (reports, exports, list views, dashboards), and you, as the party configuring the prompt and controlling the record, are responsible for how such values are configured, used and shared downstream, including the review duty above and, where they concern individuals, §2(5).

Where you use AI Features to interact with or produce content for people outside your organisation (for example AI-assisted emails at scale, or content shared with your customers): (a) where a person interacts directly with an AI system, you must disclose to them that they are interacting with AI, as many jurisdictions require this by law (for example the EU AI Act and several US state chatbot-disclosure laws); (b) the human-review and disclosure duties of §2(5) of this Addendum apply to advice, recommendations or decisions affecting individuals; and (c) you are responsible for your users' and recipient-facing deployments' compliance with this Addendum. Your end users and recipients do not need to accept any separate terms from Coevera or the model providers.

For the AI Features, Coevera acts as a deployer/integrator of third-party AI systems and is not a general-purpose AI model provider. The duties in §§2 to 4 are written functionally so that following them satisfies the common requirements of AI-specific laws in the jurisdictions where our customers operate, including the EU AI Act and US federal and state AI laws (for example AI-interaction disclosure, human review of consequential decisions, and prohibited-practice restrictions). Depending on how and where you use the AI Features, you may have your own obligations under the AI laws applicable to you. In particular, using AI Features in connection with employment-related or other consequential decisions about individuals may trigger heightened duties (in the EU, high-risk deployer obligations under Annex III; in several US states, automated-decision and employment-AI laws). A short summary of the EU AI Act position is provided in "EU AI Act summary" below; EU/GDPR matters generally are covered on our GDPR page. We can provide reasonable DPIA/FRIA assistance on request.

AI Features are disabled by default. Your administrator can enable or disable them at any time, directly from your admin settings. There is no need to contact Coevera to turn them on or off. AI Features are grouped into two packages, and both are off until an administrator turns them on:

• Voyager 1 (free AI features) is included with every subscription but is disabled by default. When enabled it consumes no credits.
• Voyager 2 (credit-based AI features) is also disabled by default and must be explicitly enabled by an administrator in the Automation Hub, with per-role controls and hard spending caps (see "The Voyager packages" below).

Each package can be turned on or off independently. Disabling a package switches off its features for your organisation; some features cannot be removed individually without affecting related functionality, so the control operates at the package and per-role level. Individual generative actions are additionally user-initiated through clearly labelled controls ("Generate", "Summarise", "AI Draft" and similar). If you would like help understanding what disabling a package affects, contact privacy@coevera.com.

Voyager AI runs inside Coevera's own AWS environment. Your data is not sent to an outside AI service such as ChatGPT or Google over the public internet. Foundation models are accessed through Amazon Bedrock (managed multi-model inference; primary model is Anthropic Claude, alongside Amazon Nova/Titan, Meta Llama, and OpenAI open-weight models contracted through AWS), with AWS Comprehend (text analysis) and Langfuse providing technical observability of the inference layer. The model providers behind Amazon Bedrock are in AWS's contractual chain, not direct Coevera sub-processors. Specific model versions and routing may change as we maintain the Services. Our use of AWS is described on our Security page, and the current list of sub-processors is maintained within our Global Terms and Services.

Processing of personal data within AI Features is governed by our Privacy Policy and, for EU/EEA customers, our GDPR commitments. In addition:

• Never sent to an outside AI service. Voyager AI does not ship your accounts, deals, emails or contacts to ChatGPT, Google or any public AI service. Inference runs inside Coevera's own secured AWS environment.
• No model training. Your inputs and outputs are never used to train or fine-tune any AI model, neither Coevera's nor those of the foundation-model providers, who are contractually committed not to use your data to train or improve their models. AWS does not retain Bedrock or Comprehend inputs/outputs for service improvement, and the foundation models themselves retain nothing after a request is served.
• Where AI processing happens (today). AI inference currently runs in AWS US East (N. Virginia, us-east-1) for all customers, including European customers. Your CRM data at rest stays in the AWS region you select (we do not move customer data between regions, see Security); when AI Features are used, the AI request is sent to and processed in the US. EU-resident AI inference (AWS EU Bedrock) is planned but is not yet available; until it is, European customers who do not want US processing of their data should leave AI Features disabled (they are off by default, see §6).
• We do keep operational records of AI requests. To run the service reliably, Coevera retains the prompts your users send to AI Features and the corresponding AI responses in our observability system (Langfuse), currently hosted in the United States. We use these records only to troubleshoot systematic issues and to improve our prompts and built-in skills so the AI Features work as intended, not to train or fine-tune any model. These records are held under Coevera's own controls and are governed by our Privacy Policy. A dedicated EU-resident Langfuse instance will be rolled out alongside EU AWS Bedrock; until then, these operational records are held in the United States. This operational retention currently cannot be opted out of at the account level; if it is a concern for your organisation, contact privacy@coevera.com.

We may suspend AI Features (not your CRM service as a whole) for your organisation or specific users where we reasonably believe this Addendum or an incorporated provider policy is being violated, or where a provider or applicable law requires it; we will give notice and restore access when the issue is resolved, where curable. We may update this Addendum; material changes take effect no earlier than 30 days after notice to your administrators, except changes required by law or by an upstream provider, which may take effect sooner with notice. Continued use of AI Features after the effective date constitutes acceptance; if you do not accept, you may disable AI Features per §6.

If this Addendum conflicts with the Terms and Conditions or the Privacy Policy, those documents prevail except on subject matter specific to AI Features, where this Addendum prevails.

Coevera's AI capabilities (branded Voyager AI) are delivered as 16 capabilities across two packages. The split determines what is free, what consumes credits, and how each is enabled. For the current, detailed breakdown of every feature in each package, see the Voyager AI Business Model & Strategy paper.

Credit-consuming features are marked with a coin icon in the product so cost is always clear. Voyager 2 enablement, credit allowance and caps are managed by your administrator; this Addendum governs how all Voyager AI Features may be used, regardless of package.

Is my CRM data used to train AI models?

No. Your data is never used to train or fine-tune any AI model, not Coevera's, and not the foundation-model providers', who are contractually committed not to use it for training. The foundation models retain nothing after answering, and AWS does not retain Bedrock/Comprehend inputs or outputs for service improvement. This is reflected in the "Data never used to train AI" commitment on our Security page.

To keep the service running reliably, Coevera does retain a record of AI prompts and responses in our observability system (Langfuse). We use these only to troubleshoot systematic issues and to improve our prompts and built-in skills, never to train models. Coevera retains these records under its own controls, currently hosted in the United States (a dedicated EU-resident Langfuse instance is planned alongside EU AWS Bedrock). See §8 ("Your data") for detail.

Where does AI processing happen?

AI inference currently runs in AWS US East (N. Virginia) for all customers, including European customers. Your CRM data at rest stays in the AWS region you select, and we do not move customer data between regions (see Security). When AI Features are used, the AI request is processed in the US. EU-resident AI inference is planned but not yet available. Because AI is off by default, European customers who do not want US processing of their data can simply leave AI Features disabled.

Which AI models and providers are used?

AI runs through AWS Bedrock and AWS Comprehend. The primary model family is Anthropic Claude; Amazon Nova/Titan, Meta Llama and OpenAI open-weight models are also used for specific features. Model providers sit behind AWS contractually. They are not direct Coevera sub-processors. The current sub-processor list is maintained within our Global Terms and Services.

Are AI features on by default?

No. Both packages are disabled by default. Voyager 1 (free) and Voyager 2 (credit-based) are each turned on by an administrator; Voyager 2 is enabled in the Automation Hub. Once a package is on, each generative action is still user-initiated via clearly labelled controls.

Can we turn AI features on or off ourselves?

Yes. Your administrator can enable or disable Voyager 1 and Voyager 2 at any time from your admin settings, with no need to contact Coevera. Voyager 2 is managed in the Automation Hub with per-role controls. Some features cannot be removed individually without affecting related functionality, so control operates at the package level. For help understanding what disabling a package affects, contact privacy@coevera.com.

How do credits work for Voyager 2?

Voyager 1 is always free and consumes no credits. Voyager 2 is credit-based and includes a free monthly allowance that resets each month, with the remaining balance shown in the Customer Portal. A hard overage limit protects you: when it is reached, Voyager 2 stops and your administrators are notified, so there are no surprise charges. Enablement, allowances and caps are controlled by your administrator in the Automation Hub.

Who owns AI-generated outputs?

You do. Outputs become part of your CRM data, like any other record you maintain.

How reliable are AI outputs?

AI outputs can be false, incomplete, misleading or out of date. They are drafts for human review: a person must check them before you rely on or share them, and for decisions that significantly affect individuals a qualified human reviewer is mandatory (Addendum §2(5)).

Do our end users or customers need to accept any third-party AI terms?

No. Your administrator's acceptance of the Addendum covers your organisation. You only need to (a) make the accuracy notice known to your users and (b) disclose AI involvement where Addendum §4 requires it. No one outside your organisation signs anything.

What happens when these terms or the providers' policies change?

Material changes take effect no earlier than 30 days after notice to your administrators. Changes forced by law or upstream providers may apply sooner, with notice. You can always disable AI Features instead of accepting.

Do AI laws (EU AI Act, US state AI laws) apply to us as a customer?

Possibly, depending on where you operate and how you use the AI Features. Most CRM use needs no extra steps beyond this Addendum. Its duties are designed so that following them satisfies the common requirements (AI disclosure, human review of consequential decisions, prohibited practices). If you use AI Features for employment-related or other consequential decisions about individuals, heightened obligations may fall on you: in the EU as a high-risk deployer, and in several US states under automated-decision and employment-AI laws. See the EU AI Act summary below and take advice for your jurisdiction.

Coevera's role

Coevera does not develop, train or provide any general-purpose AI model. We consume third-party AI services through managed gateways (AWS Bedrock multi-model inference, AWS Comprehend NLP, and Langfuse for observability). Under the Act, Coevera is a deployer/integrator of those AI systems; the general-purpose AI model obligations (Art. 53) apply to the model providers behind AWS, not to Coevera. We pass relevant upstream documentation on to customers.

Transparency (Art. 50)

Each AI feature is launched from a clearly labelled control ("Generate", "Summarise", "AI Draft"), so a user's interaction with AI is self-initiated and obvious from context. We also add voluntary "AI-generated" markers on discrete generative artefacts (generated images, AI-drafted text, AI summaries). Machine-readable marking of synthetic content under Art. 50(2) is primarily the obligation of the foundation-model provider, which we track through the AWS contractual chain. This marking does not extend to AI-assisted values that populate your structured CRM fields, which become your own business data once generated.

Prohibited practices (Art. 5)

Coevera has reviewed its AI integrations against the EU AI Act's prohibited-practice categories (in force since 2 February 2025) and does not operate any of them. In particular, Coevera's AI features do not perform emotion recognition of natural persons; the only Art. 5(1)(f) consideration is AWS Comprehend sentiment analysis on customer-directed text, which is not used to infer the emotional state of specific individuals. An internal Art. 5 review record is maintained and reviewed at least annually.

If you use AI for employment or other consequential decisions (Annex III)

If you deploy Coevera AI features in connection with employment decisions (for example evaluating sales-rep performance, allocating work, or making decisions about individuals using AI-assisted CRM data), you may be deploying a high-risk AI system as a deployer under Annex III. In that case you may need to: apply human oversight; use the system within its intended purpose; keep logs (minimum six months); monitor for anomalies and notify the provider of incidents (Art. 26); conduct a Fundamental Rights Impact Assessment where applicable (Art. 27); register the system where Art. 49 applies; maintain documentation; and comply with national employment-law overlays. We can provide reasonable DPIA/FRIA assistance on request. Take advice for your jurisdiction.

Human oversight and accuracy

All AI-generated outputs are intended for human review and must not be acted on as definitive records without verification. No AI system used by Coevera takes legal or similarly significant decisions about individuals without human intervention exercising independent discretion.

Key EU AI Act dates (reference)

Some 2026 to 2027 dates reflect staggering under the EU "Digital Omnibus" simplification initiative and remain subject to formal adoption. We track both original and deferred dates internally.

Last update: June 19th, 2026