Security
What Portward does with your credentials, your data and your results
How each is stored, what is copied from your sources, how long it is kept, and what is not in place yet.
01Every plan
Protections included in every plan
Sign-in on every request
Each request is tied to a signed-in person or a registered service. What the request says about who is asking is ignored.
Permission checks
Permissions come from your organization. An AI tool can ask for anything, and receives only what that person may see.
Rules on each source
A rule set on a source applies wherever that source is used. If Portward can't enforce a rule on a request, it refuses the request.
Encrypted credentials
Credentials are encrypted before they are stored and erased when you revoke them. They are never written to logs or errors, and never sent to an AI tool.
A record of each delivery
Who asked, what was checked and what was sent. Records are append-only and chained, so a changed or missing record can be detected.
02Storage
Where your credentials, data and results go
Credentials
Encrypted before they are stored, and never stored as plain text. Revoking a credential erases it. An AI tool never receives one.
Data from your sources
Portward copies the tables a question uses into its own storage, up to 100,000 rows per table by default and never more than 1,000,000, and runs the calculation in an isolated worker. Tables no question has used are not copied.
What can run
Only the kinds of read operation Portward supports, with checked parameters. It does not run arbitrary code or arbitrary queries on protected data.
What the AI tool receives
The data the rules allow, with its definitions. An excluded row or column is removed before anything is sent.
Definitions and approvals
An approved definition can't be edited. A correction is saved as a new version, and each approval is tied to the exact text that was reviewed and to the reviewer's sign-in.
AI models
Connecting a source and reading its structure make no AI model call. Portward runs AI processing of its own only on a paid plan, and only after you opt in.
03Retention
How long Portward keeps things
These are product defaults. They are not a compliance guarantee.
Delivery records
400 days by default. An organization can change this, down to a minimum of 30 days.
History of your requests
90 days by default.
Stored results
The planned default is 30 days. In the beta, stored results and copied tables are not yet deleted automatically.
04Status
What is in place today, and what is not
- In place
Access to your sources is read-only. Portward does not write to them, even when a credential would allow it.
- In place
The five protections above are part of every plan.
- Not yet
Automatic deletion of stored results after 30 days, and a way to request deletion yourself.
- Not yet
A published policy on which Portward staff can access customer data.
- Not yet
A third-party security certification or an independent security review.
- Not yet
Customer references. Portward is in beta.
05Questions
Security questions
Can an AI tool talk its way past a rule?
No. The AI tool can ask for anything. What it receives depends on the signed-in person, the tool, the data, the operation and where the result is going. Nothing written in a request changes who is asking or what they may see.
What does Portward see of my conversation?
Only the requests your AI tool sends to Portward and what Portward sends back. The rest of the conversation stays in the AI tool.
Can anyone see a stored credential?
A person you have authorized can display a stored credential in their browser. Every time that happens it is recorded, without the credential itself.
Can I take my data out?
Yes. Every plan includes export of the material you have the right to export.
Connect one source and ask Claude about it
Create an account, connect a PostgreSQL database, and add Portward to your AI tool.