Enable AI Parameters
Important
Before enabling this feature, you must verify that your corporate IT and security departments have approved the use of AI functionality within the Syniti Knowledge Platform (SKP).
Consultants and third-party Vendors should NOT be enabling AI without the permission of the Tenant Owner (the Customer).
Disabling AI parameters disables the use of AI functionality across the tenant.
When the Enable AI features toggle is turned on, a confirmation dialog guides you to accept the AI Terms and Conditions. If you are a company employee with approval from your corporate IT and security departments, click Accept to agree to the terms and conditions.
If you are a consultant or a third-party service provider, click the Download AI Agreement button to download an agreement form to be approved by an authorized representative of the company.
Note
This feature will be enabled for tenants individually over time. If you do not see this page in the SKP Admin module, it is not yet available for your tenant. Early access requests are available.
How the SKP uses Amazon Bedrock (Claude)
The SKP integrates Amazon Bedrock to access Anthropic’s Claude models for select AI-powered capabilities. Each request to Claude is sent independently with only the necessary, request-specific context (e.g., current record fields, query text). We do not persist prompts or outputs beyond what’s needed for the feature to function, and we do not use customer data to train any foundation models. Amazon Bedrock itself does not store or log prompts and completions by default, and it does not use your inputs to train AWS or third-party models.
Additionally, Amazon Bedrock provides strong encryption and isolation controls: TLS for data in transit and AWS KMS-backed encryption for data at rest.
Your data stays private to your tenant, is not shared with other customers or model providers, and is not used to train any foundation model.
Where the SKP uses Claude
Golden Record AI in Match Review
Helps data stewards during entity matching by intelligently recommending the most complete and accurate set of attributes for the Golden Record based on the entire match group context.
How it works:
The service sends only the fields needed for the current review task (e.g., attributes from candidate records) to Claude via Amazon Bedrock.
The response is used transiently to render suggestions; we do not store full prompts/outputs outside your tenant’s normal audit trail (if enabled).
Data Handling:
There is no training or cross-tenant sharing. Bedrock does not store prompts/completions or use them to train models. AWS controls enforce encryption in transit.
AI-assisted Web App Builder in Construct Preview
Accelerate creation of Construct web apps by turning natural language specifications into scaffolding (pages, forms, basic logic).
How it works:
We pass the current project’s spec and schema context to Claude for code scaffolding suggestions.
Generated assets are saved inside your tenant; prompts/outputs are not shared with other tenants or model providers.
Data handling:
The same Bedrock privacy guarantees apply; AWS controls enforce encryption in transit, and no data is used for training of the models.
SQL Assistant in the Migrate SQL Editor
Help users write, optimize, and understand SQL commands written against the SKP Datastores.
How it works:
The assistant sends the current SQL text and minimal table metadata necessary to generate the suggestion.
Outputs are used to populate the editor;
The command history is stored individually for each user to provide a conversation history.
Data Handling:
Prompts are not used for model training and are not distributed to third parties; encryption controls apply. Each user’s conversation history is private, as the data is stored in local cache—not accessible to any other user or service.
Security & Privacy
No retention of prompts/completions by Bedrock (by default)—Bedrock doesn’t store or log prompts and completions, and doesn’t use them to train AWS models or distribute them to third parties.
Encryption in transit and at rest—Requests use TLS 1.2+.
Tenant isolation—Bedrock enforces service-level isolation; model providers operate in model deployment accounts that they cannot access for customer content. Your data is logically isolated from other tenants.
SKP’s Data Handling Commitments
Per-request, minimal context—We send only what the model needs for the current operation; no bulk datasets or unrelated PII are included. AWS recommends avoiding sensitive data in tags/free-form fields.
No model training with your data—We never use your prompts, completions, or tenant data to train any LLM. Bedrock likewise does not use your inputs for training.
Tenant-scoped storage—Outputs that we persist (e.g., generated app scaffolding, accepted SQL) are stored inside your tenant under your policies. No cross-tenant sharing. Bedrock supports strong isolation and regional residency.
Encryption & private networking—We use TLS for requests; and, where applicable, KMS-encrypted stores.
Compliance & Governance Notes
Bedrock integrates with AWS CloudTrail for API activity, supports KMS for key management, and follows the AWS shared responsibility model.
Private connectivity (VPC/PrivateLink) and encryption controls support common enterprise requirements; AWS Bedrock has alignment with major frameworks (e.g., SOC/ISO/GDPR/HIPAA/FedRAMP), though any specific requirements should be addressed individually by tenant, as these frameworks will vary by region.
FAQ
Is any of our data used to improve Claude or any AWS/third-party models?
No. Bedrock does not use your prompts/completions to train QWS models, nor does it distribute them to third parties. We do not use your data to train any foundation model.
Does Bedrock store what we send?
No. By default, Bedrock does not store or log prompts/completions.
How is data protected in transit and at rest?
TLS 1.2+ is required for API and console traffic. For at-rest protection of applicable resources (agent configs, knowledge base ingestions, etc.), we use AWS KMS with managed keys.
Can we keep traffic off the public internet?
Yes. SKP uses VPC endpoints/PrivateLink to access Bedrock privately from within the SKP Tenant landscape.
How do multi-tenant controls work?
Bedrock enforces isolation and uses model deployment accounts per provider, which model vendors cannot access.
Contact & Support
If you have questions about how your tenant’s AI features are configured (e.g., VPC endpoints, KMS keys, invocation logging), please reach out to your Customer Success Manager or our Security & Compliance team at Syniti Support.
References
Amazon Bedrock — Data Protection (privacy, non‑retention, non‑training, model deployment accounts) [docs.aws.amazon.com]
Amazon Bedrock — Data Encryption (TLS, KMS at rest, encryptable resources) [docs.aws.amazon.com]
AWS Security Maturity Model (GenAI data protection with Bedrock) (PrivateLink/VPC, “no learning from your data”) [maturitymo…ty.aws.dev]
Bedrock Model Invocation Logging (API/config; disabled by default; destinations CloudWatch/S3; KMS encryption recommended) [github.com], [docs.aws.amazon.com], [support.icompaas.com]
Tenant Isolation Guidance with Bedrock Agents (keep tenant context in deterministic components) [becoration.com]