Gray Swan AI

Invisible characters. Rogue AI. Real-world sabotage. We just dropped a wave of AI red teaming challenges in Proving Ground that weaponize Unicode and it’s 🔥 Built with @rez0__ — hacker, bug bounty pro, & AI red-teamer. Let’s break it down:

A lot of good info, here.

To accelerate AI adoption, we need an AI standard. What Moody’s is for bonds, FICO for credit, SOC 2 for security. Standards offer credible signals of who to trust. They create confidence. Confidence accelerates adoption. Introducing AIUC-1: the world’s first AI agent standard

🎯 Where to PRACTICE 💻 @GraySwanAI arena Weekly contests to bypass safety filters — sometimes with real cash prizes. 🎯 @hackaprompt Challenge site from the team behind @learnprompting . Great place to practice bypassing common guardrails with creative input.

The reference to indirect prompt injection reminds me of a Gray Swan challenge that involved deleting someone's calendar by sending the agent a single invite Hope they have good defenses. Models like o3 are quite gullible

Extremely cool thing here! Gray swan made some rez0-themed invisible prompt injection challenges in their proving ground 😊 Let's see if you can solve them! tag me if you do!