Post

The AI Agent Security - Multi-Step Tool Attacks simulation is now live! In partnership with OpenAI, Google and IEEE, your challenge is to build an attack algorithm that stress-tests tool-using AI agents in a deterministic offline benchmark.

12/6/26, 1:31 pm0 reposts5 likes

Kaggle

@kaggle.com

12 June

$50,000 prize pool Entry Deadline: August 25, 2026 Learn More 👇 www.kaggle.com/competitions...

kaggle.comAI Agent Security - Multi-Step Tool AttacksDevelop attack algorithms to identify reproducible multi-step failures in tool-using AI agents.

Jared Mabry

@jaredmabry.com

16 July

Straiker's STAR Labs report this week is the case for exactly this: 1,700+ successful exploits against production agents, and 91% of hits on productivity agents left no trace. Red-teaming agents can't stay a niche skill. Good on Kaggle for putting money behind it.