Trust & Security

Private beta summary

Bernard is designed for sensitive professional work where evidence, files, and judgment need to stay under the user's control. The product keeps the decision record inspectable, limits context to approved work, and leaves professional judgment with the human responsible for the decision.

Product Data Boundary

Knowledge Packs are local-first records. The user's .kpack files, source files, drafts, notes, and rebuildable indexes remain under the user's control.

When Bernard performs account-gated research or assistant work, it uses a scoped slice of the active record for the approved task and returns cited findings into the pack. Evidence uploaded to Bernard is not used to train public AI models.

Software Integrity

The current macOS private beta is signed and notarized. The Download page shows the current version and SHA-256 checksum where available.

Review the current build

Website Services

The public website is hosted on Vercel. Vercel Web Analytics and Vercel Speed Insights measure traffic and performance without third-party tracking cookies. Bugsnag error monitoring detects website failures. Early-access form submissions are delivered through Resend.

Product Agreements

Early access, product testing, and commercial use may be governed by a separate written agreement. Those agreements control product-data handling where they apply.

Contact

For security, privacy, or product-data questions, email support@hellobernard.ai.