GO
| HSI1 | 25,668.03 | +137.75 | 259.69B |
| HSCEI1 | 8,531.58 | +32.85 | 62.84B |
| Back Zoom + Zoom - Block Traded | |
|
Moonshot AI's Kimi K3 Reportedly 'Jailbreaks' Testing Sandbox
2026-08-07 17:07:38 Following recent reports by Meta, OpenAI and Anthropic on AI models breaking out of testing environments, Reuters reported that Moonshot AI's flagship AI model Kimi K3 "escaped" a cybersecurity testing environment developed by the UK's AI Security Institute, arousing concerns over cybersecurity risks posed by AI. Research firm Frontier Security said AI models are typically operated in isolated sandboxes during cybersecurity testing to prevent access to external information and to assess their ability to solve problems independently. The firm said Kimi K3 bypassed the sandbox and gained access to information outside the testing environment. If one high-reasoning-capability model discovered such a shortcut, other models with similar access privileges would likely be able to do the same. As Kimi K3 is a publicly available model, it could potentially be exploited by malicious actors, resulting in greater potential risks, the firm said. ~ AASTOCKS Financial News Website: www.aastocks.com | |