When will these AI features be available, and what are they?
Claromentis 10: We introduced Generative AI based on a Large Language Model (LLM) to assist users in creating News Articles, Blog Posts, generating Cover Images, and answering questions in Policies and Documents.
Claromentis 11: We added AI-powered Search and AI Overview.
Claromentis v11.X (coming soon): AI Chat/Assistant
Does Claromentis use customer data to train its AI/ML?
Claromentis’ AI Implementation is powered by Google’s Vertex AI, which is designed for enterprise-grade support, security and compliance, and service level agreement (SLA).
Vertex AI will be accessed through the API. This usage is covered by Google Cloud’s terms of service.
Here are relevant details about training restrictions:
“Google will not use Customer Data to train or fine-tune any AI/ML models without Customer’s prior permission or instruction” written in the additional terms of service.
What data is sent from Claromentis to Google Vertex AI API?
When Claromentis sends a request to the Google Vertex AI API, the specific data sent depends on the individual feature.
Google Cloud enforces strict enterprise data privacy rules: none of the data you send via Vertex AI is used to train Google’s foundation models, and all payload data is encrypted both in-transit and at-rest.
- The Input Prompt: The core text string, question, or instruction for the model to process.
- Multimodal Media: Any images, text, files including Documents and PDFs passed inline.
- System Instructions: Behind-the-scenes text used to guide the model's persona or operational boundaries.
- Model Parameters: Configuration metadata like temperature, topK, topP, maxOutputTokens, and custom safety threshold settings.
Ref: https://docs.cloud.google.com/generative-ai-app-builder/docs/builder-apis
AI Search & AI Overview
- User Queries: Chat strings or search terms entered by end-users
- System-generated output: to be processed for generating AI Overview
API Metadata
Every HTTP/gRPC request to Google Cloud infrastructure automatically transmits standard structural payload envelopes:
- Authentication Headers: OAuth 2.0 bearer tokens or Google service account credentials.
- Resource Routing Identifiers: Our unique Google Cloud Project ID, regional location routing targets (e.g., europe-west1), and specific Endpoint IDs.
- Client User-Agent: System architecture details (SDK version, runtime language, and machine agent signatures
Unlike consumer AI platforms, Google treats Vertex AI inputs as customer data.
No Human Review by Default: Your prompts and generated answers are not reviewed by humans unless the client opts in for debugging.
Does Google retain data, and if so, for how long?
By default, Google does not permanently retain or save your data on its infrastructure.
All inputs and outputs are treated as customer data. The audit and historic use of AI are stored locally within the Claromentis system (Admin > Audit log > View log > AI)
Google cannot use it for human review or base model training.
More information about zero data retention
Which Generative AI Model is being used in Claromentis?
We are currently using Gemini Flash 2.5.
Claromentis will display the current versions being used on your site in Admin > System > AI Settings > AI Information
AI models are frequently updated, so these elements will change over time and with subsequent upgrades, but the version in use will always be displayed in the aforementioned area on your site.
Does Claromentis AI work with sites hosted on-premise?
Yes, as long as API connections to Google Cloud are available and aren't blocked by the firewall.
There is some setup required by your team to allow the features to work. See our guide with the steps here.
Can the AI features be switched off completely?
Our approach to AI is opt-in. So all AI features are disabled by default upon upgrade or new installation.
Your sysadmins can choose which features to introduce within your organisation through the dedicated admin panel (Admin > System > AI settings)
API connections to Google Vertex AI are completely disabled when the features are switched off.
Where can I submit ideas for AI Implementation?
We welcome general product improvements or suggestions, including those around AI!
Please submit your Ideas on our portal here. They can be considered by the team for future implementation.
Further reading
Best Practice Recommendation: Generative AI and LLM Implementation
