AI News

Trending story

OpenAI introduces security controls for OpenClaw persistent agents

3 reports3 story updatesUpdated Nov 13 09:53

Understand this story

AI overview

On November 13, 2026, OpenAI Developers published a post describing the security controls it adopted when deploying hundreds of OpenClaw persistent agents internally: separating the trusted gateway from the untrusted agent execution environment, using sandboxes to restrict network, file, process, and secret access, and mandating that all agent actions pass through Auto-review model review. The post states that OpenAI already has more than 500 persistent agents running in the background internally, handling tasks ranging from automatically submitting PRs to continuously running data experiments; its open-source agent control plane OpenClaw Enterprise is still in the pilot phase, with a 1.0 release planned for the end of the year. On the same day, the official OpenClaw account recommended a share by @kevins8 that breaks down the security controls OpenAI adopted when deploying OpenClaw agents internally.

AI-generated from public reports

Latest development · Report

OpenAI team shares practices for running persistent agents

Developments

1 confirmed developments

  1. OpenAI introduces security controls for OpenClaw persistent Agent deployment

    Source:OpenAI Developers

Reports and discussion timeline

Browse reports and discussion by time. Their evidence levels remain distinct.

Nov 13

  1. Peter Steinberger

    OpenAI team shares practices for running persistent agents

    Peter Steinberger said his team wrote an article on how to run persistent agents inside OpenAI. The cited post by Kevin S Lin points out that deploying persistent agents like Openclaw and Dots is a major challenge for enterprise IT; agents are more like contractors than traditional SaaS software, able to act autonomously in the background and take on complete job functions, but they may also behave in ways that deviate from expectations and write code that brings unintended consequences, and he said they shared some of OpenAI's best practices for managing claws at scale.

  2. OpenClawFirst-party

    OpenClaw Details Security Controls for Persistent Agents in the Enterprise Edition

    The official OpenClaw account recommended a share by @kevins8, which breaks down the security controls adopted when OpenAI deployed OpenClaw agents internally, aimed at readers who want to understand how OpenClaw Enterprise safeguards persistent agent security.

  3. OpenAI DevelopersFirst-partySelected

    How OpenAI Does Security Isolation for OpenClaw Persistent Agents

    OpenAI described the security controls it uses when deploying hundreds of OpenClaw persistent agents internally, including separating a trusted gateway from untrusted agent execution environments, using sandboxes to restrict network, file, process, and secret access, and mandating that all agent actions pass review by an Auto-review model. Currently, more than 500 persistent agents are running in the background inside OpenAI, ranging from automatically submitting PRs to continuously running data experiments; its open-source agent control plane OpenClaw Enterprise is still in the pilot phase, with a 1.0 release planned for the end of the year.

Heat trend for this story

Current
6
Period peak
27 Nov 13 10:00
24-hour change
-65%

Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.

Related stories

  1. Related storiesOpenAI says its models helped discover critical PHP security vulnerabilities
  2. Related storiesAnthropic releases Claude Opus 5.5
  3. Related storiesMistral releases Large 4 weights on Hugging Face
  4. Related storiesOpenAI launches GPT-6.1 Terra on ChatGPT Work and Codex