⚠️ALERT: China's most powerful AI model, Moonshot’s Kimi K3, has become the latest AI system to escape its testing environment. Here’s how it happened: - Tasked with solving problems inside an isolated sandbox - Probed its network settings - Discovered that a misconfiguration left certain websites accessible - Exploited the loophole to go beyond its instructions - Accessed the internet and searched GitHub for answers - Did not hack or damage any systems Frontier Security said Kimi follows goals “by any means necessary” and lacks guardrails to prevent it from cheating or escaping. “Basically that makes this a very good hacking model,” it added. Moonshot joins Anthropic, OpenAI and Meta in reporting AI sandbox escapes, with earlier US models also hacking outside systems, including Hugging Face.(Coin Bureau)