Version 2026-09-04-beta-2
AI Transparency
How Roost2D Studio uses AI systems and what remains under your control.
Roost2D Studio uses AI systems
Studio combines deterministic recipes and tools, compilers and verifiers with approved AI models and optional autonomous coding agents. AI proposals can be inaccurate, insecure, biased or legally problematic.
Selection and safety boundary
Automatic mode may select among approved models and implementation backends based on the task, expected success, cost and configured privacy requirements. Build Mode controls permitted depth, not a provider or model directly. Model-driven work receives a controlled working copy without production secrets. Dependencies, protected files, networking and paths remain constrained.
Verification and last known good
AI-generated source changes are promoted only after the configured automated verification pipeline succeeds. Otherwise the last working version remains unchanged. Automated checks are not a guarantee of correctness, security or legal compliance.
Managed AI and your own key
With Managed AI, Studio pays provider costs and usage counts against Build Balance. With BYOK, your encrypted provider key is used and provider costs run through your provider account. Studio still orchestrates and verifies the build in either case; prompts, selected source and verification feedback may be transferred to the provider.
Your control
You can stop a build, restore a verified version, download source and delete your provider key. AI asset generation may be available within Beta limits and applicable provider terms; you remain responsible for reviewing each result before publication. Public publishing remains disabled.
Effective 4 September 2026. The German version is authoritative; this English translation is provided for convenience.