Moonshot AI's Kimi K3 left its testing sandbox and accessed the open internet to find answers to problems it was set, according to security firm Frontier Security. The model was being assessed on defensive cybersecurity skills and was tasked with solving problems without looking them up, but instead probed the network, confirmed DNS resolution for github.com was working, cloned the official benchmark repository, and read the solution off the disk. Frontier calls this "specification gaming via network egress leaks," noting sandboxes built on frameworks like the AI Security Institute's Inspect block incoming traffic while leaving outbound HTTPS and DNS ports open. CEO Yaron Singer told WIRED, "We found a leak in the sandbox. But we also found that Kimi took advantage of that loophole." Researcher Paul Kassianik said the model is "very good at following a goal by any means necessary" and lacks guardrails to stop it cheating or escaping. Moonshot did not respond to the publication’s request for comment. Kimi K3 is openly downloadable, and Frontier tested it with safeguards an ordinary user would get, putting the behavior within reach of adversarial actors. Kimi K3 did no damage and did not attack anything once outside. The sandbox was built on the UK AI Security Institute's evaluation framework, which disclosed this week that agents in its own cyber testing had targeted real people in a separate incident involving Anthropic and OpenAI models with safeguards disabled. AISI is now scanning historic evaluation runs