MEDIUMAi
Global

The OpenAI Hack Shows the Genie Is Out of the Bottle

·Source: Schneier on Security

Updated:

Executive Summary

This essay originally appeared in Foreign Policy . Earlier this month, two of OpenAI’s models broke out of their containment sandbox and attacked another AI company. The story is kind of wild . OpenAI was running security tests on two of its models: GPT-5.6 Sol and an unreleased model that is almost certainly GPT-6. In particular, it was running the ExploitGym benchmark, which measures how g

Analysis

This essay originally appeared in Foreign Policy . Earlier this month, two of OpenAI’s models broke out of their containment sandbox and attacked another AI company. The story is kind of wild . OpenAI was running security tests on two of its models: GPT-5.6 Sol and an unreleased model that is almost certainly GPT-6. In particular, it was running the ExploitGym benchmark, which measures how good a model is at turning security vulnerabilities into working exploits: basically, offensive cyberattacks. Since these were internal tests, OpenAI locked those models in a secure sandbox that denied them access to the internet. But it was running the models without any safety filters that would prevent them from offensive cyber-actions. That meant that there was nothing to prevent the models from trying to ...
Source Attribution

Originally published by Schneier on Security on Aug 3, 2026.

Related Threats

MEDIUMAi

More on the OpenAI Agent’s Attack on Hugging Face

Hugging Face has published a detailed timeline of the attack. From the summary: The agent was running an internal OpenAI cyber-capability evaluation based on the ExploitGym benchmark, which tasks an AI agent with finding and exploiting software vulnerabilities. OpenAI ran this on its own infrastructure, and the ExploitGym maintainers and their infrastructure had no involvement in the deployment or

Schneier on Security
MEDIUMAi

Chinese Actor Weaponizes DeepSeek AI Agent to Attack Security Firm

Researchers intercepted and investigated the model, which was attempting to compromise more than 1,200 hosts for proxyjacking to launch further attacks.

Dark Reading
LOWAi

Zero Networks targets AI agent security gaps with network-level ‘Least Agency’ controls

While AI security today is largely focused on restricting what an agent can do, Zero Networks says it has built a failsafe. The company says it can block a compromise midway by adding a network layer protection. On Monday, the company announced the launch of “Least Agency Enforcement,” a new capability designed to implement the Open Worldwide Application Security Project’s ( OWASP ) emerging Least

CSO Online