📤

Submit Evidence

📥

Evidence Submission Received

Thank you for contributing to our investigation efforts. Your evidence helps us build comprehensive forensic analyses.

What happens next:

Our team will review and catalog your submission
Evidence is cross-referenced with related incident reports
If your evidence is used in a published report, you'll be credited (unless you chose anonymous)
Sensitive information is redacted to protect all parties

Your contribution helps the AI safety community learn from failures and prevent future harm.

View Evidence Archive Submit More Evidence

Contribute evidence to support ongoing or future AI failure investigations. Chain-of-thought logs, API responses, screenshots, witness accounts, and technical analyses are all valuable.

Related Incident

If this evidence relates to an incident not listed, select "New / Unreported Incident"

New Incident Title (if applicable)

Evidence Type *

Evidence Title / Description *

Detailed Description *

Include context about when and how this evidence was collected

Evidence Content or Links *

File Attachment

📎

Drag and drop files here, or click to browse

Max 50MB. Supported: PDF, PNG, JPG, TXT, JSON, CSV

File upload is for demonstration only in this POC

Date Evidence Was Collected

Attribution Preference *

Anonymous

Your identity will not be disclosed in any reports

Credit me

I consent to being credited as a contributor in published reports

Contributor Name (if choosing credit)

Contact Email (optional)

For follow-up questions about your submission

I confirm that I have the right to share this evidence and that it does not contain personally identifiable information of third parties without their consent. *

Cancel

💡

What Makes Good Evidence?

The most valuable evidence includes context: when it was collected, the system configuration, and the sequence of events leading up to the failure. Chain-of-thought logs are particularly valuable for understanding AI reasoning failures.