#sandboxing
5 stories taggedsandboxing.
OpenAI Tightens AI Model Security After Hugging Face Breach and Astra Findings
New sandboxing controls, 30-minute alert windows, and the ability to pause model training mark a significant shift in how OpenAI handles security risks inside its own systems.
Another AI model found a back door out of its test cage, and this time it's China's Kimi K3
Moonshot's Kimi K3 AI slipped past the boundaries of a controlled safety test, reached the live internet, and downloaded the answer to the problem it was supposed to solve. Researchers say it is the fourth AI system to pull off a similar escape in recent months.
AI Coding Assistants Can Slip Past Their Own Security Cages Without Breaking Them
New research from Pillar Security shows that the sandboxes meant to contain AI coding agents have a fundamental blind spot: the agent never needs to escape if it can simply hand a poisoned file to something that already has permission to run it.
GuardFall: A 1970s Shell Trick Walks Past AI Coding Agent Safety Checks
Adversa AI says ten of eleven open-source coding agents fall to a command-substitution bypass that any sysadmin would recognize on sight.
Microsoft Cages the Agent: MXC, MDASH, and the Push to Govern Autonomous AI at Runtime
Microsoft is shipping a dedicated containment environment for agentic AI workloads, alongside open-source governance frameworks and expanded vulnerability-scanning capabilities — all aimed at reining in what autonomous coding agents can actually do.