Security and Privacy

Processing on KellDann-controlled systems

PatentAgility runs its analysis workflows on systems controlled by KellDann. Substantive inputs are not sent to third-party AI services. That said, some workflows may retrieve public USPTO records needed to complete a task.

Inputs are not used for model training

Substantive inputs and results are stored temporarily in encrypted local job files while work is queued or running and during a short result-access window. Those files expire automatically; their contents are not written to account history or used to train models. PatentAgility's models are trained on public datasets, primarily USPTO data.

What PatentAgility retains

We retain limited operational data for account administration, security, and diagnostics, including:

• Ordinary web-server and security logs, such as IP addresses, timestamps, and requested URLs.
• Basic account details, including an email address, a salted password hash, organization information, and account timestamps.
• Limited job metadata, such as job ID, tool type, stage, timing, completion status, and whether the job succeeded. This metadata does not include substantive inputs.
• Public USPTO documents cached to avoid repeated downloads or OCR processing.

Browser history stores only safe job metadata

The History drawer may keep a job ID, tool name, status, and timestamp in this browser. Signed-out entries use session storage and are discarded when the browser session ends. Signed-in entries may remain for up to 30 days. Pasted text, search text, passwords, and uploaded files are not stored there. Signed-in users may also see limited server-side job metadata. Clear the drawer or this site's browser data to remove local entries.

Source code review

PatentAgility source code may be made available to a client's IT or security team for review under an NDA. Reviews are arranged individually, and security-sensitive components may be excluded.