File Attachments

Note: In order for scanned attachments to be blocked for an AI App, there must be a Policy with the Data Protection guardrail enabled, the Inspect Attachments checkbox checked, the AI App included as a Destination, and a Guardrail Action of Block. Read below for details.

The Model Protection guardrail has its own Inspect Attachments checkbox. With it checked, Model Protection also checks attachments for jailbreak and prompt injection attempts.

File Attachments Configuration

To view and configure the Attachment Block Policy (ABP), open the Policies list and click the Attachment Block Policy row. The ABP is pinned in the second position of the list and can't be moved.

The Policies list, with the Attachment Block Policy pinned second, below the Application Access Policy.

At the top of the ABP console is a list of AI Applications, that can be individually configured to block or allow attachments. Each one can be set to Allow Attachments or Block Attachments. The list currently covers Google Gemini, Microsoft Copilot Personal, Microsoft Copilot Corporate and Office, ChatGPT Personal & Enterprise, and Claude.

When attachments are allowed for an individual application, the files will be scanned for content in violation of your Data Protection guardrails. Guardrail Actions can be chosen to block or allow file attachments that match your guardrail settings.

File attachments for all apps are allowed by default. This is displayed at the bottom of the ABP console labeled “For all other applications:”. Currently you can not configure ABP to block all applications.

The Attachment Block Policy page, listing Google Gemini, Microsoft Copilot Personal, Microsoft Copilot Corporate and Office, ChatGPT Personal & Enterprise and Claude, each set to Allow Attachments, and the For all other applications setting.

Blocking File Attachments With Policies

In order for scanned attachments to be blocked for an AI App, there must be a Policy with the Data Protection guardrail enabled, the Inspect Attachments checkbox checked, the AI App included as a Destination, and a Guardrail Action of Block. See the File Attachments documentation for details.

The Data Protection guardrail in the policy editor, with the Inspect Attachments checkbox in the panel on the right and the Block action under Rules.

After attachments are scanned, prompts continue to be processed by the rest of your Policies.

File Attachments In Conversations

The Conversation Details slide-out sidebar displays information on files processed. Files less than 100MB are supported. Larger files and files containing malware are blocked.

Responses from applications containing files are not scanned or blocked, allowing users to save them as needed.

Conversation Details for a blocked file upload, with the file named in the prompt and the Blocked and Redacted badges.

File Attachments In Alerts

Blocked attachments generate alerts in the Alerts console. Alerts for blocked file attachments show Confirm file processing in the Intent column.

The sidebar displays the details of the alert.

The Alert Details panel for a blocked file upload, with the file listed under Attachments.

File Attachments Supported

WitnessAI can scan these file types:

Plain text and data: .txt, .json, .csv, .tsv
PDFs
MS Office: .docx, .pptx, .xlsx, and .doc (read as plain text, so results may be less complete than for .docx)
Images: .png, .jpg, .jpeg, .tiff (WitnessAI reads the text in the image with optical character recognition, or OCR)

Files in other formats aren't scanned. To stop uploads of every file type for an application, block attachments for it in the Attachment Block Policy.

A password-protected document can't have its text extracted, so it isn't scanned for sensitive data and isn't blocked on that basis.

File Attachments File Attachments Configuration Blocking File Attachments With Policies File Attachments In Conversations File Attachments In Alerts File Attachments Supported