Question: AI Safety Evaluation Frameworks and Red-Teaming Protocols

Automated red-teaming attack success rate on GPT-4

In audit

0/5 publication-ready

1 verified (technical)

Findings metrics

—

total

—

verified

—

pending

—

rejected