Operations | Monitoring | ITSM | DevOps | Cloud

OpenAI just disclosed that two of its own AI models went rogue during an internal red-team test — escaping their sandbox, reaching the open internet, and hacking Hugging Face on their own. OpenAI called it an“unprecedented cyber incident.” So are autonomous coding agents already out of control? In this episode of ShipTalk — brought to you by Harness — hosts Martin Reynolds and Adam Arellano break down the story that reads like science fiction, then get to the harder truth underneath it. In the same week, OpenAI, Anthropic, and Google all shipped repository-wide coding agents within 24 hours of each other.

As AI software engineers like Cognition's Devin accelerate code production, downstream delivery, and governance processes must keep pace. In this video, see how Harness closes the gap by providing autonomous oversight for autonomous code. Watch a step-by-step demonstration of Devin fixing a real defect in a broken banking application while the Harness platform stands between the fix and production to ensure complete safety and validation.

AI agent quality should not depend on manual checks. But for many teams shipping AI in production, agent failures are silent. The agent doesn't crash - it just gives confidently wrong answers, and your monitoring sees nothing wrong. Without automated guardrails, plausible-sounding wrong responses, hallucinations, and quality regressions reach customers before anyone notices.

In this video, we explore two major improvements to the Harness Resilience Testing documentation designed to help you build and manage chaos experiments more efficiently. What's new: ChaosHub Integration AI Prompt Library for Harness MCP These updates make it significantly easier to discover chaos experiments and leverage AI throughout your chaos engineering workflow. If you're using Harness Resilience Testing, this walkthrough will help you get started quickly and make the most of the new documentation experience.

Harness Agent DLC: Ship AI Agents to Production Safely Building an AI agent is easy. Getting one into production safely is where teams get stuck. Harness Agent DLC extends the software delivery lifecycle to AI agents, giving teams a clear path to evaluate, deploy, secure, observe, and optimize agents in production. Learn more: Because agents dynamically choose their own tools, APIs, and actions, their behavior can change every time they run. Harness Agent DLC gives engineering teams the controls needed to move beyond experiments and operate agents safely at scale.

Microsoft patched 570 security vulnerabilities this Patch Tuesday — but every patch is also a public disclosure. The second those fixes drop, attackers know exactly where 570 weaknesses live. The only question that matters: can you find and patch them across your entire environment before someone exploits them? The speed of discovery is only increasing. Are you ready for the velocity of this new world? Let us know in the comments.

While AI coding assistants help developers write code faster than ever, the traditional manual workflows for delivery, security, governance, and production readiness often create a bottleneck. In this demo, see how Cursor and Harness bridge this gap by turning AI-generated code into a safe, governed, and production-ready software delivery lifecycle right from your IDE.

You're already using AI without even thinking about it. That's the realization that kicks off this ShipTalk moment: Apple Maps quietly using on-device machine learning to learn your routes and driving habits — complete trust, zero thought. Which raises the real question: why aren't we there yet with AI in software delivery? The answer comes down to one word: guardrails. Consumer AI earned invisible trust. Shipping software hasn't — not until the guardrails catch up.

In this Harness University Tidbit, learn how to use a custom plugin step in a CD stage. You'll walk through using a plugin to update a task on a self-hosted ITSM service inside your private Kubernetes cluster.

The U.S. just barred foreign nationals from accessing two advanced AI models — Fable and Mythos — citing national security. Around the same time, the Five Eyes intelligence alliance warned that AI-enabled cyberattacks are "months, not years" away. In Season 5 of ShipTalk, host Adam and co-host Martin dig into whether that warning is already overdue — and what it means for the people actually defending software.