AI in CloudVision360 and ITOps360 follows the same rule as every other part of the platform: it works for people, under policy, and it leaves a record.
What our AI will never do
- Invent facts about your environment. Answers are computed from your tenant’s data and cite the findings, resources and policies they rely on.
- Claim that a change succeeded before it has been verified.
- Run a privileged action without permission, or bypass an approval policy.
- Approve a change. Only people can approve.
- Reveal secrets or another tenant’s data.
- Use tools outside the allow-list for its feature, or escalate its own privileges.
How you control it
- On or off, per tenant. AI features can be disabled entirely for your tenant.
- Your data stays with us. In our hosted service, customer data is never sent to an external AI provider. AI features use only self-hosted models; until one is available they are off and the products give their standard answers.
- Private deployments. When you run the products in your own environment, you choose the provider: your own self-hosted model, or a provider such as Azure OpenAI, Amazon Bedrock, Google Vertex AI, Anthropic or an OpenAI-compatible API, with the highest data classification that may reach it.
- Budgets and quotas. Monthly budgets, per-user daily limits, per-feature quotas and rate limits.
- Transparency. Every request records the prompt version, a hash of the prompt, the provider and model, tokens and cost.
How we build it
- Prompts are versioned and reviewed like code.
- AI features fall back to deterministic answers when AI is unavailable, so nothing depends on a model being right.
- Prompt-injection and tool-abuse cases are part of our test suites, and AI behavior is in scope for penetration testing.
- We align our AI governance with the NIST AI Risk Management Framework and are preparing for ISO/IEC 42001.