[Wave of AI Security Vulnerabilities Spreads: Two Risks Exposed in Muse, Plans to Strengthen Risk Alerts] According to a report by Jinse Finance, on September 26, security vulnerabilities were discovered in Meta Platforms' AI product Muse, which could potentially affect user data and virtual environment security. In response, Meta plans to enhance risk alerts. It is reported that the vulnerability was submitted by an external security researcher through Meta's bug bounty program and had not been publicly disclosed before. The vulnerability allows attackers to access users' dedicated virtual machines, obtaining cloud-based personal account data such as emails and documents. A Meta spokesperson revealed that the vulnerability was initially classified internally as SEV-2, the second-highest level in the company's five-tier security grading system, typically corresponding to security incidents with significant impact. After further evaluation, it was downgraded to SEV-3. Meta stated that exploiting this vulnerability requires certain preconditions: users must either aggregate information in Muse or click on a malicious webpage link and then select "Allow" in a security pop-up to grant authorization. To address this risk, Meta will introduce more prominent warning alerts. If Muse detects that a user is about to access a malicious website, it will display an enhanced warning as an additional protective measure.