X · by Aditya Rao

Catching colluding agents by reading their chat

Jev named the colluders 85% of the time.

Aditya Rao

@dev_adityarao

Catching AI agents that secretly collude usually means reading their minds. I just read their chat instead, using Jev by TypeSafe AI It named the colluders 85% of the time. Built for the AI Swarm Dynamics Hackathon by @grove_research & @aidigest_. Thanks for letting people join online, I really enjoyed it!

Oct 5, 2026 · 4 likesOpen on X

Aditya Rao detects AI agents that secretly collude. The usual approach reads the model's internals; he reads their chat with Jev instead. He reports that it named the colluders 85% of the time. He built it for the AI Swarm Dynamics Hackathon.

Open the source

More like this

Hassan
@nutlope
Jev + Kimi K3 for fraud detection! TLDR: Jev classified 100 emails in 1.42 seconds, then I routed the uncertain cases to Kimi K3. The full pipeline got 96/100 correct for only ~$0.07. Video is not sped up, check out the live run! Here was my process: I gave Jev 100 emails to classify (a mix of 50 legit & 50 fraudelent emails). It classified all of them in 1.42 seconds. An underrated feature about Jev is it will give you the confidence score for a classification, so I routed any prediction under 95% confidence to Kimi K3 to be fully sure. 31 emails fell below that threshold. After routing those to Kimi K3, the combined pipeline reached 96% accuracy. The full run took 16 seconds & ~$0.07 in inference costs: - $0.068 from Kimi K3 on @togethercompute - $0.003 (1/3 of a cent) from Jev on @typesafeai. I think this is a really interesting pattern: use a fast specialized model like Jev for the narrow task, then route the uncertain cases to a larger LLM. I feel like this kind of approach could be a game changer for use cases like fraud or anything realtime. You can use the speed & low cost of Jev while having a larger LLM as a fallback to ensure high accuracy.
XSecurity and abuse
Pick
Fraud detection with Jev and Kimi K3
Emails
100 in 1.42 s
Correct
96/100
Cost
~$0.07 pipeline ($0.003 Jev, $0.068 Kimi K3)
😎Nick 常胜
@isNickMa
Tested TypeSafe’s Jev (no-text, probability-only model) as an AI agent safety monitor. Checking each action first worked well caught most attacks with almost no false blocks, and much faster than Gemini.
XSecurity and abuse
Jev as an agent safety monitor
Frank Chen
@francchen
I don’t know if people still remember Jev. Things move so fast here. I’ve spent the last few days testing it, and found a few things I think builders should see. Prompt injection is one of the most interesting things to test in AI, so I made a little demo to show how it could change Jev’s answer. Jev is a great model. I just want people to know what to watch out for when they use it.
XSecurity and abuse
Testing Jev for prompt injection

Catching colluding agents by reading their chat

More like this

Fraud detection with Jev and Kimi K3

Jev as an agent safety monitor

Testing Jev for prompt injection