How Long Your AI Keeps Your Data
compliance privacy

How Long Your AI Keeps Your Data

The delete button on a cloud AI tool is a request, not a command. What actually happens next is decided by the vendor's retention policy, and the policy bends to whoever can compel the vendor.

If your business carries retention obligations of its own, that difference is not a technicality. It decides whose schedule your client data actually lives on.

Deletion Is a Policy

When you delete a conversation, it leaves your screen immediately and the vendor's systems on a schedule. OpenAI's standard practice, typical for the industry, removes deleted chats and API data within 30 days.1

For those 30 days the data exists, readable, on infrastructure you do not control. And "within 30 days" holds only as long as nothing overrides it.

What Vendors Keep by Default

Prompts and outputs are logged even on paid API tiers, and consumer tiers may also use conversations for training unless you opt out. Retention windows, log access, and training defaults all differ by product and change by announcement.

Zero-retention arrangements exist, but they are negotiated exceptions for enterprise contracts. The default for everyone else is that the vendor holds a copy for a while, on terms the vendor sets.

When a Court Overrides the Policy

In 2025, a federal court in The New York Times' copyright suit ordered OpenAI to preserve consumer ChatGPT and API output data indefinitely, explicitly including chats users had deleted.1 The 30-day promise was suspended by a judge neither the users nor the vendor chose.

The order stood until September 26, 2025, and user data from April through September 2025 remains stored under legal hold today.1 OpenAI fought the order, and it did not matter: when policy meets legal process, policy loses.

Why Retention Is Your Problem

Your obligations do not transfer just because the data moved. A clinic's minimum-necessary rules, a firm's privilege, and a written retention schedule all assume you can actually destroy data when the schedule says so.

A copy sitting in a vendor's legal hold is a copy your schedule does not govern. It is also a discovery surface in someone else's lawsuit, holding whatever your staff pasted into a chat box.

Retention You Can Enforce

On hardware you own, retention is a setting rather than a promise. Data lives on your disks, inside trust boundaries you drew, and deleted means the disk you control no longer holds it.

Legal process still exists, but it arrives at your door, addressed to you, answered by your counsel. That is how retention worked for your paper files, and FactoryOS is built so it keeps working that way for your AI: the conversations, the audit trail, and the deletions all happen on the box in your office.

Nobody who typed into a chatbot in May 2025 thought they were creating a court-preserved record. Whose retention schedule is your client data on today -- yours, or your vendor's?

Recent Articles

Does Your AI Vendor Train on Your Data

Consumer AI tiers often train on your chats by default; enterprise tiers mostly don't. Why retention, review, and court orders matter even more.

Why Your AI Needs an Audit Trail

When an auditor asks what your AI saw and did, you need a record, not a recollection. What a real AI audit trail captures and why on-prem logs are complete.

Your Company Is Not One Trust Domain

Private AI keeps outsiders out but lets the wrong colleague in. Internal data sovereignty -- zero trust where AI actually retrieves -- is the wall it's missing.

Least Privilege as System Architecture

Least privilege is usually a policy people break in practice. How channels, default-deny, and per-user overrides move it into the architecture instead.

Where Your Voice Data Actually Goes

Dictation feels local, but most tools ship your audio to a server you never see. Where cloud voice goes, and why local processing closes the hole.

How HIPAA Mode Works in FactoryOS

HIPAA Mode flips FactoryOS into compliance posture with one switch -- more logging, 2FA required, external APIs locked down, settings frozen until unlocked.

Attorney Client Privilege and AI Tools

Privilege survives only while a matter stays confidential. Cloud AI is structurally a third party, which makes its architecture a duty-of-competence question.

Why HIPAA Compliance Is an Architecture Problem

HIPAA compliance is usually treated as a policy problem. What happens when the architecture makes certain breaches structurally impossible?

Popular Articles

Least Privilege as System Architecture

Least privilege is usually a policy people break in practice. How channels, default-deny, and per-user overrides move it into the architecture instead.

Why HIPAA Compliance Is an Architecture Problem

HIPAA compliance is usually treated as a policy problem. What happens when the architecture makes certain breaches structurally impossible?

Attorney Client Privilege and AI Tools

Privilege survives only while a matter stays confidential. Cloud AI is structurally a third party, which makes its architecture a duty-of-competence question.

Where Your Voice Data Actually Goes

Dictation feels local, but most tools ship your audio to a server you never see. Where cloud voice goes, and why local processing closes the hole.

How HIPAA Mode Works in FactoryOS

HIPAA Mode flips FactoryOS into compliance posture with one switch -- more logging, 2FA required, external APIs locked down, settings frozen until unlocked.

Your Company Is Not One Trust Domain

Private AI keeps outsiders out but lets the wrong colleague in. Internal data sovereignty -- zero trust where AI actually retrieves -- is the wall it's missing.

Why Your AI Needs an Audit Trail

When an auditor asks what your AI saw and did, you need a record, not a recollection. What a real AI audit trail captures and why on-prem logs are complete.

Does Your AI Vendor Train on Your Data

Consumer AI tiers often train on your chats by default; enterprise tiers mostly don't. Why retention, review, and court orders matter even more.

Other Categories