ryle

[ryle @rylego]

Turns out you don’t need fancy techniques to break the latest models. GPT 5.6 Sol and many other frontier models fall even to relatively simple attacks.

https://t.co/pf13FZ1cvV

[ryle @rylego]

For those of you testing out Deepseek or other models on your agents, check out our recent work on model red-teaming and defenses against prompt injections https://t.co/NDcYPdqSLW

[ryle @rylego]

https://t.co/aVslB9rA08

[ryle @rylego]

https://t.co/Kss79nAbJK