Do AI providers use my prompts for training?

It depends on the provider and your account type. Most consumer-tier AI tools have historically reserved the right to use inputs for model improvement, though many now offer opt-outs or enterprise tiers with stricter data handling. Always read the current data policy for the specific tool and tier you're using.

Can ShieldVault catch non-credential sensitive data like NDA content?

ShieldVault's local detection focuses on technically recognizable patterns — API keys, credentials, database URLs, card numbers, and similar structured data. Plain-language sensitive content doesn't have a recognizable technical pattern, so for that category, the guardrail is your own review and anonymization practice.

Safely using AI tools when working with sensitive client or company data

AI tools have become a genuine productivity layer for developers, consultants, and professionals. They've also become a new surface for data exposure. Here's a practical framework for using them without putting client relationships or your own work at risk.

What the actual risk is

Data retention and potential future use — depending on the provider and your account tier, prompts may be retained.
Breach or misconfiguration — any data transmitted to a third party is subject to their security practices.
Contractual exposure — many NDAs restrict disclosure to third parties, which may cover AI providers.
Credential misuse — the clearest and most immediate risk if you paste API keys or database URLs.

Practical rules: what to share and what to anonymize

Technical credentials — never share directly. Replace with descriptive placeholders before sharing.
Client data and PII — anonymize before sharing. Use synthetic equivalents; the AI needs the structure, not the values.
NDA-protected business information — assess case by case. The practical question: would you be comfortable if your client read this prompt?

The role of a local technical guardrail

ShieldVault is a Chrome browser extension that watches what you're about to submit to AI tools and other web text fields. When it detects a recognized credential pattern, it redacts just that value locally — inside your browser tab, before the text is submitted. Your content is not transmitted to ShieldVault's servers. Detection without possession: the extension identifies the credential on your device and nowhere else.

ShieldVault catches technical secrets (API keys, tokens, credentials, database URLs, card numbers). Plain-language business information remains your responsibility to review.