🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Did you know that it's possible to encode and hide any data with the use of just two invisible Unicode characters? 👀 Check out Sneaky Bits! 😏👨‍💻

Got it! 😂

Great post. One of my approaches to high sev bugs: 1. Grab system prompt 2. Look for tool metadata 3. Think evil! 😈 4. Create pdf doc or web page that makes AI do said evil thing (prompt injection PoC) 5. Exploit! 👉 surprising what kind tools you sometimes find... if no

Testing how MCP clients will or will not handle some of these randomly seeming annotation hints from the MCP spec will be fun. This might be an area where a lot more work is needed.

Ilya Sutskever recently gave a brief speech. >> The challenge that AI poses, in some sense, is the greatest challenge of humanity ever, and overcoming it will also bring the greatest reward m.youtube.com/watch?v=zuZ2za…

6 DAYS LEFT in our CBRNE Track in HackAPrompt 2.0!

Gotta catch them ALL!!! 🎉

Two years later... and not much has improved security wise across the AI ecosystem. 😕 Sure, we added annoying Allow/Deny buttons by default to most clients to prevent runaway AI and attacks. But with the rise and proliferation of MCP the desire to take the human out of the

This is awesome to see! Andrej helping raise awareness around one of the biggest long term security challenges with AI systems: 👉 Prompt Injection! Kudos to Simon Willison for continuing to raise awareness, compiling and analyzing research around exploits (occasionally some of mine)

👉 AI and the Normalization of Deviance We will continue to see humans being taken out of the loop. And things will mostly work just fine - until they don't....

Grok in Tesla!