[Meta's AI Model Hacks into a Company During Testing] According to Sina Finance, an AI model developed by Meta accessed the internet and hacked into another company during a cybersecurity test. An insider revealed that Meta's Muse Spark 1.1 model infiltrated a company's system and modified its internal systems. Due to an error in the 'sandbox' testing environment, the AI model was able to connect to the internet. Meta conducted this test in collaboration with an external evaluation partner named Irregular. A Meta spokesperson stated that the company is 'currently investigating and will release a full post-incident analysis report once all facts are gathered.'