FAQ
Does PromptGlass decrypt all of our web traffic?
Section titled “Does PromptGlass decrypt all of our web traffic?”No. The gateway decrypts TLS only for domains in the AI app catalog plus any domains you add. All other traffic is tunneled without inspection. You can also list domains that must never be inspected, and that list takes precedence. See gateway forward proxy.
Do we need all four sensors?
Section titled “Do we need all four sensors?”No. Most organizations start with the browser extension and one or two connectors, add the gateway for API traffic and inline control, and add the endpoint agent for desktop and IDE coverage.
Can PromptGlass block data before it reaches the AI tool?
Section titled “Can PromptGlass block data before it reaches the AI tool?”Yes, where it sits inline: the browser extension and the gateway can coach, redact or block before a prompt is sent. Connectors read vendor logs after the fact, so they alert instead of block.
Does it add latency?
Section titled “Does it add latency?”Inline detection runs in the gateway process. Pattern and dictionary detectors take well under a millisecond. ML classifiers add more and only run when a rule needs them. You set a detection time budget (detectors.timeoutMs), and on_error decides what happens if it is exceeded. AI model response times are usually much longer than the detection overhead.
Will employees know they are monitored?
Section titled “Will employees know they are monitored?”They should. PromptGlass includes a transparency page and an extension popup that show what is captured. Pseudonymization and four-eyes re-identification are on by default. See employee privacy.
Does PromptGlass train AI models on our prompts?
Section titled “Does PromptGlass train AI models on our prompts?”No. Your data is never used to train PromptGlass models or anyone else’s. Custom classifiers you build are trained inside your tenant and used only there.
Which AI tools are supported?
Section titled “Which AI tools are supported?”Any web or API-based AI tool can be monitored through the gateway and extension, using a continuously updated AI app catalog. Deep integrations (API connectors) exist for Microsoft 365 Copilot, ChatGPT Enterprise and the OpenAI API, Claude and the Anthropic API, GitHub Copilot, Gemini for Workspace, Azure OpenAI, AWS Bedrock and Vertex AI.
Can we run it entirely on-premises?
Section titled “Can we run it entirely on-premises?”Yes, on any Kubernetes 1.29+ cluster, including fully air-gapped. See on-premises and air-gapped.
Does PromptGlass make us compliant with the EU AI Act or ISO/IEC 42001?
Section titled “Does PromptGlass make us compliant with the EU AI Act or ISO/IEC 42001?”No tool makes you compliant on its own. PromptGlass helps you produce evidence: an inventory of AI systems in use, usage records, policy enforcement logs and AI literacy signals. Those support your obligations under frameworks such as the EU AI Act, ISO/IEC 42001 and the NIST AI RMF.
How is it priced?
Section titled “How is it priced?”Per monitored employee per month, billed annually, starting with a free 30-day AI usage assessment. Self-hosted and air-gapped deployment is available on the Enterprise plan. See pricing or contact us.