■■□□□ Thread: CrowdStrike published a blog at the beginning of the month, exposing new prompt injection techniques. This time, 18 new injection techniques have been added, bringing the project’s total coverage to over 200 different injection techniques. The CrowdStrike AI security research team claims to maintain the industry’s largest-scale prompt injection classification system, providing a structured hierarchical framework that clearly demonstrates the full spectrum of risks posed by artificial intelligence threats.
https://x.com/i/status/2081582153884930237