Frequently Asked Questions
How does TrustPrompt work?
TrustPrompt detects sensitive information in your text (names, companies, emails, account numbers, etc.) and replaces them with placeholder tokens like [PERSON_A] or [COMPANY_B]. The sanitized prompt is then sent to the AI. When the AI responds, TrustPrompt restores the original values in the response.
Does my sensitive data leave my browser?
By default, no. Detection runs locally in your browser using on-device machine learning. The mapping between your real data and placeholder tokens stays in your browser. Only the sanitized version (with placeholders) is sent to AI providers. If you enable "Deep Analysis," your text is sent to a third-party AI for enhanced detection before tokenization. We use providers with zero data retention policies where available, but we cannot independently verify their data handling. For maximum privacy, keep Deep Analysis off.
What's the difference between Managed Mode and BYOK?
Managed Mode: Use our API keys. You pay with credits. Simple setup, no API keys needed. BYOK (Bring Your Own Key): Use your own API keys from OpenAI, Anthropic, etc. Your browser connects directly to the provider. You pay them directly based on their pricing.
How do credits work?
Credits are used in Managed Mode to pay for AI responses. Each AI model has different credit costs based on token usage. Sanitization itself costs 0.5-1 credit depending on whether Deep Analysis is enabled. Subscription plans include monthly credits that roll over (up to 2x your monthly amount). For example, if your plan includes 100 credits/month, you can accumulate up to 200 credits.
Is TrustPrompt 100% accurate?
No automated system is perfect. TrustPrompt reduces the risk of exposing sensitive information, but detection is probabilistic and may not catch everything automatically. That's why you're always in control. Before sending, you can review what's been detected and make adjustments: add protection for missed items, remove protection for incorrectly flagged terms, and save patterns to teach TrustPrompt for the future. Think of TrustPrompt as a safety net that catches most sensitive data automatically, with you making the final call on what gets protected.
Does TrustPrompt help with compliance (HIPAA, GDPR, etc.)?
TrustPrompt can help reduce the risk of exposing sensitive data in your AI workflows. However, it is not certified for HIPAA, GDPR, or other regulatory compliance. Detection is probabilistic—always review the sanitized output before sending. If you work with regulated data, consult your compliance team before use.
What types of sensitive data does TrustPrompt detect?
TrustPrompt detects: names, companies/organizations, email addresses, phone numbers, physical addresses, account numbers, social security numbers, credit card numbers, dates, monetary amounts, and more. Detection improves with Deep Analysis enabled.
Can I use TrustPrompt with any AI?
In Managed Mode, we currently support Claude (Anthropic), GPT-4 (OpenAI), and other popular models. In BYOK mode, you can use any provider whose API key you have.
Who built TrustPrompt?
TrustPrompt is built by Pulse Insights LLC, a company on a mission to turn listening into impact—helping organizations understand and act on customer needs through AI-powered tools.