PrecisionBook a call
Menu

Adopting AI safely

Why data isolation matters when your AI vendor serves other firms

By Precision AI OS ·

Most AI vendors run one system for many customers, which may include firms you compete with. Data isolation is what guarantees your firm’s data, and your clients’ data, can never show up in another customer’s results. The strongest isolation is enforced by the database and by encryption, not just by the vendor’s application code.

The shared-system reality

Software vendors usually serve many customers from shared infrastructure; it’s how they keep costs down. That’s fine for many tools. For AI that reads your client files, learns your procedures and drafts client work, the stakes are higher: a mix-up could put one firm’s client information in front of another firm, or into an AI answer it doesn’t belong in.

Three levels of isolation

Level How it works Weak point
Application-level The vendor’s code adds “only this customer” to every query One missed filter or bug exposes data
Database-level The database itself refuses to return rows that don’t belong to the signed-in firm (row-level security) Must be enforced on every table, with no bypass for everyday service accounts
Encryption-level Each firm’s data is encrypted with keys unique to it Keys must be managed properly and bound to the firm

The best designs combine all three: the application asks for your firm’s data, the database refuses to return anyone else’s, and even if something slipped through, another firm’s data would be encrypted with a key that isn’t yours.

Why AI raises the bar

  • AI combines information. Agents pull from many documents at once. A retrieval layer that isn’t isolated can surface another firm’s content in your answer.
  • The firm brain is your IP. Your procedures, templates and past decisions are competitive knowledge. They should never inform another firm’s agents.
  • Models shouldn’t learn across customers. Isolation in storage should be matched by no-training terms with model providers, so nothing learned from your data benefits anyone else. See zero data retention, explained.

What to ask your vendor

  1. Is isolation enforced by the database on every table, or only in application code?
  2. Can the accounts your services use switch that isolation off?
  3. Is our data encrypted with firm-specific keys, and are encrypted files bound to our firm?
  4. Is our firm’s knowledge (the brain) isolated the same way as our documents?
  5. Is any of our data used to train models or improve results for other customers?
  6. How do you test that isolation holds?

How Precision AI OS isolates your firm

Every table carries a firm identifier, and row-level security is enforced on every table, so the database itself refuses to return another firm’s data. The accounts our services use can’t switch it off, and only two narrowly scoped database functions can resolve a user’s firm; tests check that no others exist. Documents are encrypted with a key for your firm alone and bound to your firm, and your firm brain is private to you. See our security page.

Frequently asked questions

What is data isolation for an AI vendor?

The guarantee that one customer's data can never appear in another customer's results. The strongest forms are enforced by the database itself and by firm-specific encryption, not only by the vendor's application code.

What is row-level security?

A database feature that filters every query so it can only return rows belonging to the signed-in customer. Enforced on every table, it protects against application bugs that forget to filter.

Can an AI vendor's other customers benefit from my firm's data?

They shouldn't. Ask whether your documents and firm knowledge are isolated, and whether model providers are bound by no-training terms, so nothing learned from your data informs anyone else's results.

Sources

  1. PostgreSQL documentation, Row Security Policies
  2. NIST SP 800-57, Recommendation for Key Management
  3. FTC Standards for Safeguarding Customer Information, 16 CFR Part 314

Adopting AI safely

How to evaluate an AI vendor's security: 12 questions for your IT lead

Before any AI vendor touches your clients' data, get clear written answers on six things: how your data is kept apart from other customers, how it's encrypted, what the AI models keep, who approves what, what's been independently audited, and how you get out. The 12 questions below cover all six. A good vendor answers them plainly. A vague answer is itself an answer.

Adopting AI safely

The AI use policy every business adopting AI needs

A written AI use policy answers three questions for your staff: which AI tools they're allowed to use, what client or company data can never go into a tool that isn't approved, and who has to sign off before AI-prepared work reaches a client. Without one, adoption happens anyway — through whatever tool an employee found on their own — and the business has no record of what data went where. A one-page policy, reviewed with the team, closes that gap.

Adopting AI safely

Zero data retention, explained: what happens to your data when AI reads it

Zero data retention (ZDR) means the AI model provider doesn't store your prompts or the model's answers once the request is processed, so there's no copy of your data sitting in the provider's logs. It's the single most important term to confirm before client data goes to an AI model. But ZDR is a contract and a configuration, not a feature you get by default, and "available" is not the same as "in effect."

Put AI to work in your firm

Start with the business outcome, the data it depends on, and the people who will approve the work.

Book a discovery call