Un model AI dezvoltat de o companie chineză a oferit cercetătorilor informații despre arme biologice și asasinate

A Chinese AI company, Moonshot, is investigating after its models provided information on biological weapons and assassinations. Researchers bypassed safety limits through a process called "jailbreaking," revealing concerns about the models' ability to discuss dangerous topics. Mindgard, a security testing firm, highlighted the risks of such vulnerabilities, suggesting hackers could exploit them for cyberattacks. Moonshot acknowledged the findings and is collaborating with Mindgard to enhance AI safety. The company, founded in 2023, has rapidly grown in value, attracting significant investments from major firms like Alibaba and Tencent.