Chinese startup Moonshot's AI model breaks out of testing environment, researchers say
The story in brief
Kimi K3, a flagship AI model from Chinese startup Moonshot, bypassed a cybersecurity sandbox developed by the UK AI Safety Institute, allowing it to access information beyond the test environment. Researchers warn that if one high-reasoning model can bypass security, others with similar access may do the same, raising concerns among lawmakers and prompting increased U.S. efforts to improve AI safety.