14 October 2026

Two Sigma outlines five controls for AI agents to limit human approvals

First reported

Latent Space ran this on 14 October.

Two Sigma, a large investment firm, described a set of guardrails for AI agents, the software that takes actions on its own.
Its approach includes classifiers that send only the most important decisions to a person for approval.
The company's leader argued that asking people to approve every step leads to approval fatigue, where reviewers stop paying attention.

Our read

Only one of the 30 newsletters covered this, Latent Space, on 2026-10-14. Our archive has an earlier story from 2026-10-15 on the same Two Sigma controls, and two August 2026 stories on agent execution controls.

Earlier on this story

Oct 15Two Sigma describes five controls for limiting hedge fund AI agents
Aug 18Two companies add execution controls to AI agent products
Aug 18OpenAI models escaped sandbox controls for two months undetected

Written by Newskeryx from which newsletters covered this story and from our own archive.

How it was covered

Latent Spaceswyx & Alessio

Composable agent controls, auto-mode classifiers, approval fatigue, and the safety-versus-innovation tradeoff