Mask prompts with coherent fictional values.
Replace PII, customer identifiers, or financial figures inline before submission via EU-hosted inference (OVHcloud Gravelines) with an 18-second fail-closed timeout.
ChatGPTMask or block sensitive prompts sent to ChatGPT.
Protect ChatGPT sessions in real time through the Elba browser extension. Automatically mask sensitive entities with coherent fictional values, block regulated prompts and text-file uploads, and separate browser governance from your OpenAI Admin API seat audit.
Managed browser extension · Separate OpenAI Admin API connector for seats
Capabilities
What this integration covers.
Block regulated prompts and file uploads.
Stop source code, credentials, and sensitive text or code uploads (up to 5 files, 16,000 characters) from leaving the browser.
Distinguish personal ChatGPT from API seats.
See work vs. personal browser accounts on chatgpt.com while auditing developer API memberships separately via the OpenAI Admin API connector.
Example workflow
Sanitize or block a sensitive ChatGPT prompt in flight.
How the integration supports a specific security task.
- 01
Inspect before submission
The managed Elba browser extension evaluates a ChatGPT prompt or text-file attachment before it leaves the endpoint.
- 02
Mask or block by policy
Your Playbook either rewrites sensitive entities with coherent fictional values or blocks the send outright.
Coach and log
The employee receives immediate in-browser context while your security team reviews the alert metadata in Elba.
Related capability
Explore this part of the platform.
Before you connect
Coverage, permissions, and requirements.
Inline ChatGPT prompt and upload controls run through the Elba browser extension on chatgpt.com. For paid ChatGPT Team or Enterprise seat inventory, User Access Reviews, and automated deprovisioning, Elba also provides a separate OpenAI Admin API connector (users and remediation).
Explore security & privacyHow does Automated Prompt Masking work on ChatGPT?
Before a prompt leaves the browser, Elba detects sensitive entities and rewrites them into coherent fictional equivalents so ChatGPT can still answer the question without receiving real data. If rewriting exceeds 18 seconds, Elba fails closed and blocks submission.
What is the difference between ChatGPT browser coverage and the OpenAI API connector?
The Elba browser extension governs live prompts, text uploads, and personal-account usage on chatgpt.com. The separate OpenAI Admin API connector inventories paid OpenAI seats for User Access Reviews and supports automated account removal.
Where are prompts analyzed and are raw prompts stored?
AI prompt classification runs on European infrastructure via OVHcloud AI Endpoints in Gravelines, France. Elba stores zero raw prompt bodies and never trains AI models on customer data.
Which uploaded files are inspected on ChatGPT?
Elba inspects file names for uploads and analyzes content excerpts for up to 5 supported text or source-code files (up to 16,000 characters each). Binary formats such as PDFs, Office files, images, and archives expose metadata only.
Coverage and requirements
- Browser-observed ChatGPT sessions are not verified directory accounts; use the separate OpenAI Admin API connector for seat inventory and deprovisioning.
- Text-file content inspection covers up to 5 supported text or code files per upload (up to 16,000 characters per file).
- For binary files, including PDFs, Microsoft Office documents, images, and archives, only file metadata is inspected; binary bodies are not extracted.
- Automated Prompt Masking enforces an 18-second timeout and fails closed (blocking the send) if rewriting cannot complete in time.
Other applications
Review more integrations.
Discuss your requirements
Evaluate Elba for ChatGPT.
Review the available capabilities, required permissions, and setup for your environment with our team.