🛡 Launch of Declaw Arena: A CTF Platform for Hacking AI Agents
The Declaw Arena platform has been introduced for testing the security of AI agents. Users can attempt to hack agents within isolated micro-VMs (sandboxes) with varying levels of protection: ranging from no policies to PII filtering systems.
🌍 The tool highlights the need for security testing at the runtime and environment policy level, rather than just at the prompt level.
👤 An opportunity to test AI Security skills by bypassing protection mechanisms in a controlled environment.
Source 1: https://declaw.ai/arena