AI and Data Use Policy
Version 1.0. Last updated: 18 November 2026
| Policy details | Details |
|---|---|
| Entities | CIM Environmental Pty Ltd and its subsidiaries ("CIM"): CIM Operations Pty Ltd (ABN 14 169 162 243), CIM Enviro Europe Ltd and CIM US Co |
| Applies to | The PEAK Platform, the PEAK AI Connector, AI agents and features in PEAK, CIM's machine learning models and CIM team AI tools |
| Version | 1.0, 18 November 2026 |
| Review | At least annually, and whenever CIM's use of AI changes |
| Contact | data.protection@cim.io |
1. Summary
CIM uses AI to help the people who run buildings find and fix problems faster. We do this without giving up control of your data. Our commitments:
- AI providers that CIM uses are not permitted to train their models on customer data.
- AI features work within PEAK permissions. They can only see or change what the signed-in user, or the agent's own account, is authorised to access.
- Customer data is stored in Amazon Web Services (AWS) data centres in Sydney, Australia, and AI agents and features in PEAK process it in Australia. The providers we use, and where they process data, are listed in section 8.
- Our own machine learning models learn from equipment and meter data. We remove customer and site identifiers before data from different customers is combined for training.
- AI agents in PEAK are shown as AI wherever they appear, every action they take is recorded, and customers can ask us to switch them off.
- The PEAK AI Connector does not collect your prompts, your AI client's responses or your conversation history. If your AI client keeps a history of your chats, it stays with your AI provider and CIM cannot see it.
- PEAK only changes building control settings, such as setpoints, with the customer's explicit permission, and every change is logged.
- We do not sell Customer Data or personal information.
2. Scope and how this policy fits with our terms
This policy explains how CIM uses AI and how the AI features of the PEAK Platform ("PEAK"), CIM's building analytics software, handle customer data and personal information. It is also the privacy notice for the PEAK AI Connector.
This policy sits alongside, and does not change, our Standard Licence Terms and Conditions (Terms), your Service Agreement and our Privacy Policy. If they differ, your agreement prevails.
AI output can be incomplete or wrong, so check it before acting on it. Our Terms apply to AI features as they do to the rest of PEAK, including the warranty and high-risk use terms in clauses 9 and 10.4 and the ban on getting around access controls in clause 7.2.
In this policy:
Table 1: Definitions
| Term | Meaning |
|---|---|
| CIM, we, us | CIM Environmental Pty Ltd and its subsidiaries, including CIM Operations Pty Ltd (ABN 14 169 162 243), CIM Enviro Europe Ltd and CIM US Co |
| Customer Data | Data that customers and their users provide to or create in PEAK |
| Machine Level Data | The building, meter and equipment data defined in clause 4 of our Terms |
| PEAK | The PEAK Platform, CIM's building analytics software |
| Terms | Our Standard Licence Terms and Conditions |
| Training | A model learning from data. Using a model to answer a request or complete a task is inference, not training |
3. How CIM uses AI
CIM uses AI in the four areas set out in Table 2.
Table 2: How CIM uses AI
| Area | What it does | Where the AI runs | Is your data used for training? |
|---|---|---|---|
| PEAK AI Connector | Lets your users query and update PEAK from their own AI client | The AI provider your organisation chooses | CIM does not train on connector data. Your AI provider's handling depends on your account with them. |
| CIM team AI tools | Helps our team support you, analyse building performance and prepare reports | Claude and ChatGPT business accounts managed by CIM | No. The providers are contractually prohibited from training on our data. |
| AI agents and features in PEAK | Automates routine work and adds AI assistance inside PEAK | Amazon Bedrock, in CIM's AWS environment | No. Customer data is not used to train third-party models. |
| CIM machine learning models | Improves commissioning, data quality checks and how PEAK analyses building data | Amazon SageMaker, in CIM's AWS environment | Yes, using Machine Level Data with customer and site identifiers removed. |
4. PEAK AI Connector
4.1 Overview
The PEAK AI Connector lets authorised PEAK users connect PEAK to an AI client, meaning an AI assistant such as Claude, ChatGPT or Microsoft Copilot, using the Model Context Protocol (MCP). Users can ask their AI client about sites, equipment, alerts, action tickets and performance data, and can create or update action tickets, comments and fault detection rules where their PEAK role allows.
4.2 Access and permissions
- Each user signs in with their own PEAK account. There are no shared keys.
- AI Connector sign-in uses PEAK's multi-factor authentication, which we recommend turning on for all users.
- The connector can only see and do what that user can already see and do in PEAK. It does not widen anyone's access.
- Anything created or changed through the connector is recorded in PEAK under the user's name, in the same way as any other edit.
- Users can disconnect at any time from their AI client, which ends the connector's access.
- A customer can ask CIM to turn off AI Connector access for its whole organisation by emailing security@cim.io.
4.3 What CIM receives
CIM receives only the requests the AI client sends to the connector, such as "search action tickets for this site", and returns the results. We do not request, receive or store your conversation history, prompts or the AI client's responses. Your AI client may keep a history of your conversations, depending on its settings. That history is held by your AI provider, and CIM has no access to it.
For each request we record the tool used, the time, the user and the parameters the AI client sends, which can include IDs and site, client, contractor or equipment names. We use these records for security monitoring and usage analytics. Text written into PEAK through the connector, such as comments and ticket descriptions, is redacted before it reaches our analytics tools.
4.4 Your AI client is your choice and your responsibility
Once PEAK returns data to your AI client, it is handled under your agreement with that AI provider, not this policy. Your organisation chooses the provider, the account type and its settings, including how long conversations are kept and whether the provider may train on them. We recommend business or enterprise accounts with model training turned off.
5. How CIM's team uses AI tools
Our team uses AI tools to support customers, for example to analyse equipment and energy performance, investigate faults, draft reports and answer support requests. Team members may use the PEAK AI Connector with their own login, or work with extracts of Customer Data in an approved tool.
- We only use CIM-managed business accounts of Claude (Anthropic) and ChatGPT (OpenAI), which contractually prohibit training on our data. Personal and free AI accounts are not allowed for work.
- Staff use only the data a task needs, and never enter credentials, security details or sensitive personal information.
- If a customer's contract prohibits sharing its data with third-party services, we do not use that customer's data in these tools.
- These rules come from CIM's internal AI Usage Policy, which every team member must accept.
6. AI agents and features in PEAK
6.1 Overview
CIM builds AI agents and other AI features into PEAK to help customers and their contractors get work done. An agent can carry out tasks on its own, such as following up outstanding work, adding context, updating records or preparing summaries, using the Customer Data it is permitted to access. The commitments in this section apply to every agent and AI feature we add.
6.2 How agents are built and hosted
Agents and AI features in PEAK run on Amazon Bedrock inside CIM's AWS environment, using models from several developers that are hosted and operated by AWS. We only use models that AWS hosts and operates, so model developers do not receive the prompts or outputs, and customer data is not used to train third-party models. Further information can be found on AWS's Bedrock security and privacy page. Customer data stays stored in AWS Sydney, Australia by default, though some models may process a request in another AWS region.
6.3 Safeguards
- Labelled as AI. Agents are shown in PEAK as AI. Content and actions created by an agent are clearly identified as coming from AI, so people know they are not dealing with a person.
- Limited permissions. An agent acts either through its own account, limited to the sites and actions it is enabled for, or on behalf of a signed-in user within that user's PEAK permissions. It cannot see or do more than that.
- Building control needs permission. Changes to building control settings, such as setpoints, need the customer's explicit permission and must be turned on or approved by an authorised person. Every change is logged.
- Recorded and reversible. Changes an agent makes in PEAK are recorded in PEAK's history and can be reversed by a person.
- Customer control. Customers can ask CIM to switch off any agent for their sites or portfolio.
- Reviewed before release. Every AI feature is approved by CIM's product and technology leadership before it goes live.
6.4 Decisions about people
Our agents automate operational work in PEAK, such as handling equipment faults, tickets and data quality. They do not make decisions about individuals that have legal or similarly significant effects. Decisions about people, such as engaging or assessing a contractor, remain with our customers and their staff.
7. CIM's machine learning models
CIM builds its own machine learning models to help PEAK analyse building data, including for commissioning. We train and run them on Amazon SageMaker in CIM's AWS environment, and they are not shared with third parties.
- We only train them on data our Terms allow us to use for this purpose. Today that is Machine Level Data (clause 4). The models and the insights derived from them are CIM Data under clause 4.4.
- Fault detection rules created by a customer are that customer's Confidential Information under clause 7.1 of our Terms. We do not train on them without the customer's agreement.
- Before data from different customers is combined for training, we remove customer and site identifiers.
- Where Machine Level Data contains personal information, we only use it as clause 8.3 of our Terms allows.
8. Personal information
This section summarises how we handle personal information under this policy and is the privacy notice for PEAK AI Connector users. Our Privacy Policy covers everything else.
Table 3: Personal information summary
| Topic | Details |
|---|---|
| What we collect | Account information (name, email address, organisation, role and PEAK user ID); usage data, including connector requests as described in section 4.3; IP address and browser or AI client details; and content users create in PEAK, including through the connector. We do not collect payment details, government identifiers, health information or precise location, and we do not receive AI conversation history through the PEAK AI Connector. |
| Why we use it | To provide PEAK and its AI features within each user's permissions; to secure the service and detect misuse; to troubleshoot problems; to understand usage and improve PEAK; and to meet legal obligations. Usage data, including connector request records, is not used to train AI models. Where the GDPR or UK GDPR applies, we rely on our contract with your organisation and our legitimate interests in securing and improving PEAK. |
| Who receives it | The providers listed in Table 4 and the service providers listed on our Trust Centre; the AI provider you connect, which receives the results you ask for; authorities where the law requires it; and a successor business if CIM is sold or restructured. |
| How long we keep it | We keep Customer Data while your organisation's account is active, and after it ends unless your organisation asks us to delete it. For 30 days after an agreement ends, your organisation can ask to export its Machine Level Data. Deletion requests are completed within 30 days, and backup copies are removed as backups expire. Data already de-identified for model training, and CIM Data under clause 4.4, are not affected. |
| Your choices | Disconnect the connector at any time, and ask CIM to switch off the connector or agents for your organisation. You can access, correct or delete your personal information as set out in our Privacy Policy by emailing data.protection@cim.io. Because PEAK is provided to organisations, we may refer requests to your organisation's administrator. |
Customer data is stored in AWS in Sydney, Australia. The providers in Table 4 process data for our AI features and AI tools, some of them outside Australia. The analytics and support tools that receive PEAK usage data, including connector usage data, are located in the United States and are listed with our other service providers on our Trust Centre.
Each provider is assessed before use and bound by contract to protect the data. Where privacy law requires a transfer mechanism, such as standard contractual clauses, we rely on the provider's data processing terms.
Table 4: AI providers
| Provider | Purpose | Location |
|---|---|---|
| Amazon Web Services | Hosting and storage, AI agents and features (Amazon Bedrock) and machine learning (Amazon SageMaker) | Australia (data stored in Sydney) |
| Anthropic (Claude) | AI tool used by CIM's team, with extracts of Customer Data needed for a task | United States and other provider locations |
| OpenAI (ChatGPT) | AI tool used by CIM's team, with extracts of Customer Data needed for a task | United States and other provider locations |
9. Security
AI features run within the same information security program as the rest of PEAK. Our Trust Centre lists our security controls, and customers can request our latest SOC 2 report, security policies and latest penetration test there. See also our Security and Compliance page.
10. Changes and contact
We may update this policy as our use of AI changes. The current version is always on this page, with the date it was last updated.
For privacy questions or requests, email data.protection@cim.io. To report a security issue, email security@cim.io. Our office addresses and phone numbers are on our contact page.
Powering property teams in these world leading companies.

