AI FAQ
Common questions, answered plainly.
AI assists with tasks like generating infrastructure code, reviewing pull requests, drafting deployment configurations, and summarising incidents. It produces suggestions for a person to review — see Human Oversight for how that review works.
AI can draft infrastructure changes and prepare a plan, but applying that change — especially in production — requires human approval through our standard deployment workflow.
Only where you've explicitly enabled a feature that needs it (for example, AI code review), and only the parts of the code relevant to that feature — not a standing, unscoped read of your entire codebase.
Yes. AI-assisted features are opt-in per feature; you can turn them off in your account or project settings without losing access to the rest of the platform.
Where we support it, yes — for example choosing between AWS Bedrock-hosted models or a provider you configure yourself. Not every feature supports provider choice today; we're expanding this over time.
Not without your explicit, opt-in consent. See Privacy & Data for the full detail.
Through the same tooling as human-written changes: terraform validate/plan, test suites, linters, security scanning, and human review — see AI Risk Management and AI Development Standards.
Support for self-hosted/local models is limited today and varies by feature. Where it's available, it's called out in that feature's configuration.