Safety and privacy

People approve every changeCopy Anchor Link

• New projects start read-only.
• Every change to a model, ACC or a shared project rule is shown first and needs a person's approval.
• Scheduled runs (for example a daily issue digest) never change anything. They leave proposals that a person approves later.
• Every statement in a report links to the evidence it's based on, so it can be checked.

The agents never see passwordsCopy Anchor Link

Logins for ACC, Microsoft 365 and other systems are kept in the computer's own secure storage (the Windows Credential Manager or macOS Keychain). The tools that connect to those systems use them; the AI agents never see them. Even if someone hides instructions in an issue or a document, there is no password to steal, and nothing changes without approval.

Confidential project dataCopy Anchor Link

• Everything stays in your own project repositories on your own git host (for example GitHub or Azure DevOps).
• Raw exports (Relatics, planning) and meeting transcripts are kept out of the shared project history by default. Only the results (minutes, reports) are shared with the project team.
• Your personal desk, which spans projects for different clients, is private to you.
• Project sites are private by default and are never published without a deliberate choice.

Visibility of work and works councilsCopy Anchor Link

bimai keeps a short journal of what each session did, so colleagues can pick up where someone left off. It describes work and outcomes, never time spent or productivity scores. Some organizations (in the Netherlands, the works council) will want to agree on this before rollout; the journal can be switched off per project.

Which AI model is usedCopy Anchor Link

bimai runs on Claude through Claude Code, under your organization's own account and data agreements. Most work runs on smaller, cheaper models; see Costs.