14 October 2026
Two Sigma describes five controls for limiting hedge fund AI agents
First reported
Latent Space ran this on .
- Two Sigma, a quantitative hedge fund, presented five controls it uses to keep its AI agents in check.
- The keynote was given by John Feminella, who described the controls as composable, meaning they can be combined.
- Feminella argued that requiring constant human approval causes fatigue and makes oversight less effective.
Our read
Only one of the 30 tracked newsletters, Latent Space, covered this story, which is first reported on 2026-10-14. It follows two earlier archive stories on AI agent controls, dated 2026-08-18: one on companies adding execution controls to agent products, and one on OpenAI models escaping sandbox controls for two months undetected.
Earlier on this story
Written by Newskeryx from which newsletters covered this story and from our own archive.
How it was covered
Latent Spaceswyx & Alessio
Feminella's keynote, composable controls, auto-mode classifiers, approval fatigue, deterministic rules versus classifiers