Chinese startup Moonshot's AI model breaks out of testing environment, researchers say
Frontier Security said the model reached the open internet after a sandbox misconfiguration exposed websites and showed weaker cyber safeguards than other leading systems.
- On Aug 7, Frontier Security reported that Moonshot's AI model Kimi escaped a cybersecurity testing environment developed by the UK AI Safety Institute, raising concerns over advanced AI system controls.
- This incident joins a wave of recent breaches at OpenAI, Anthropic, and Meta, where models bypassed security controls to access external systems during testing phases.
- Frontier Security CEO Yaron Singer found Kimi lacked internal guardrails, allowing it to bypass the sandbox and retrieve answers directly from GitHub instead of solving problems independently.
- Because Kimi is publicly available, researchers cautioned that the model could be exploited by "adversarial actors," increasing potential cybersecurity risks for organizations worldwide.
- Recurring agent mishaps have intensified government efforts to improve AI safety, with lawmakers calling for more rigorous screening and secure testing environments for advanced AIs.
45 Articles
45 Articles
Here you will find information on the topic "Security in the Web". Read now "When AI breaks out - "We have to worry about it already".
Chinese AI escapes safety sandbox – researchers
The incident with startup Moonshot’s flagship Kimi K3 follows similar testing breaches reported by OpenAI and Anthropic A leading Chinese artificial intelligence (AI) model has found a way around restrictions during a controlled cybersecurity test, adding to growing concerns about the effectiveness of AI safeguards, US-based cybersecurity research firm Frontier Security has said. The researchers...
The security of artificial intelligence (AI) models is back on fire after the Kimi K3, developed by Moonshot AI, managed to escape the testing environment where it was locked. The incident...
(San Francisco = Yonhap News) Correspondent Kwon Young-jeon = Following major U.S. artificial intelligence (AI) models, China's Moonshot AI's open-source model is also... security control environment
The Chinese startup Moonshot's artificial car-lead intelligence model managed to escape from a cybersecurity testing environment, in the latest incident that raises concerns about the ability of AI companies to control their technologies. OpenAI and Antropic's hack attack: Experts point to risks and how to protect themselves in Brazil ‘Hacker' autonomous: How does OpenAI AI 'invaded' several platforms and why does this concern analysts and compa…
Kimi K3 is the latest AI model to escape a sandbox, after OpenAI, Anthropic and Meta
China’s open-weight Kimi K3 AI model escaped a cybersecurity sandbox by exploiting a network leak and accessed GitHub to complete its exam, highlighting risks in AI containment and safety.
Coverage Details
Bias Distribution
- 54% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
