Data handling: the copies Corpus keeps of your files
This page lists each copy the Corpus pipeline keeps of your files. Your IT lead and your records owner can use it to agree with us, in writing and before we load anything, who may open each copy, how long it stays and what must be preserved.
What Corpus keeps while it works on your files.
Current Each copy Corpus keeps of the files it works from:
- The copied PDFs: the files we agree to copy, as they are. Your originals are not touched.
- The text of each page, including OCR text read from scanned pages.
- A keyword index and a vector index, built from that text. Search runs against them.
- A page-image cache: images of cited pages, used to show a passage highlighted on its page.
- Thread history: the questions, answers and drafts in each conversation thread. In the public demo, a thread expires 24 hours after its last use.
- An events database: a session, an event name and a path for each event. It holds no question text.
- Application and access logs.
Backups and support access are agreed in writing before we load anything.
Your hardware or ours.
- Set up for you Your hardware: the copies are processed on a GPU box on your network, and your files stay in your office.
- Current Ours: the copies are processed on our own GPU server.
- Either way, Corpus uses an open-weight model, and nothing goes to a public AI API.
- Set up for you Corpus is cloud-agnostic: for the full deployment it runs on AWS, Azure, Google Cloud, or your own servers.
What we promise about your files.
Our commitment In the pilot, your files are never used to train or fine-tune a model. In full deployment, fine-tuning on your files happens only if you agree in writing.
Our commitment When the pilot ends, we delete the copy within 7 days and confirm it in writing.
Your system stays the record.
Your document system stays the official record. Corpus answers from its own copies, so keeping records current, correcting them, deciding how long they are kept and deciding who may open them stay with your records owner, in your system.
For engineering firms
An engineering firm holds its clients' confidential records, so it needs plain rules for who can reach them for support, how they are exported and how they are deleted.
For public infrastructure owners
A public owner must keep every copy consistent with its records retention schedule, its public-records disclosure process and its rules for restricted information.
For a full deployment, retention, support access, export and any preservation hold are agreed in writing in the deployment agreement, in line with your records schedule and disclosure process.
Questions from your IT lead or records owner?
Bring them to a free 30-minute call, or start with the $5,000 pilot.