Privacy Policy

We collect the following categories of information:

• Account data. Email address, a generated username, a password hash (if you set a password), display name, and — if you sign in with Google or GitHub — the provider's subject identifier and the verified email they asserted.
• Project and source-code data. Specifications you type, follow-up messages, uploaded RTL project archives, uploaded PDFs for the knowledge base, generated Verilog/VHDL, testbenches, simulation logs, synthesis reports, and related run artifacts.
• Billing records. Subscription status, run charges, credit-ledger entries, and payment-provider identifiers. These are financial records and survive deleting a chat.
• Support tickets. Messages and attachments you send to support, and our replies.
• Run statistics. Token counts, model identifiers, and USD cost of a run, used for capacity, pricing, and (if you opt in) product analytics.
• Feedback you submit after a run (thumbs, reasons, ratings, and optional comments).

Uploaded project files and generated RTL are stored so a chat can be reopened and a run can be continued. They live on disk under the author-output directory for this instance, with metadata in our database. We do not use your RTL to train our own models. We do send the text of a request — and, when a run requires it, relevant source — to the model provider serving the model you selected. See Data processing.

• Chats and generated files. Until you delete the chat, or until the account is deleted.
• Project workspaces attached to a chat: removed after 30 days of inactivity. An uploaded project that was never claimed by a chat is removed after 24 hours.
• Billing and ledger rows. Kept as financial records even after a chat is deleted.
• Account. Until you ask us to delete it (currently via a support ticket).

Chats and projects are private to your account. Other customers cannot see them. Staff can see a run's files only if you explicitly opt in when submitting feedback ("Share this run with the SiliCode team for investigation"). A share link you create makes that thread readable by anyone who has the URL; it is the only other way to expose a chat. There is no team-wide visibility and no cross-account access.

SiliCode does not train models on your specifications, source, or generated RTL. Requests are sent to third-party model providers so they can generate a response. Whether a provider retains or trains on API traffic is governed by that provider's terms; we select providers whose API products are not used for training by default. Details are on the data-processing page.

We use PostHog (EU-hosted) for product analytics and fully masked session replay, so we can see where the interface confuses people. Typed text is not captured as text. A switch in Settings turns future collection off for your account. Disabling analytics does not change how runs behave, what they cost, or what you can access.

We set a session cookie that holds only a session identifier (the session itself is stored server-side). The product SPA also stores a login token in localStorage. We do not use advertising cookies or sell personal data.

• Delete a chat from the sidebar; the conversation and its generated files go with it.
• Discard a project workspace from the upload panel, or delete the chat that owns it. Unused workspaces are removed automatically as described under retention.
• To delete your account, open a support ticket from Settings → Support or email [email protected].

Traffic is encrypted in transit (HTTPS). Passwords are stored as hashes, never in plaintext. API tokens are revoked on logout. Uploaded archives are scanned for zip-slip and symlink members, and execution bits are stripped. We do not claim encryption at rest as a product feature on every host; ask us if you need that stated for a particular deployment.

Privacy questions: [email protected]. For a data-processing addendum, see Data processing.