CISA Orders Patch for Cisco Flaw Tied to AI Agent Attack

The Cybersecurity and Infrastructure Security Agency (CISA) has ordered federal agencies to patch a Cisco BroadWorks cloud internet telephone platform flaw by Nov. 24, two weeks after Malbear Labs reported that Chinese-speaking hackers used a jail-broken DeepSeek AI model to exploit the vulnerability and breach a European telecom. The researchers said an AI agent sent a rigged XML message to a BroadWorks server, stealing private keys and accessing a health-care database protected only by default credentials. The agent operated almost entirely autonomously for more than two weeks, mapping 13 organizations and writing its own replacement tool when endpoint security blocked the original. Cisco’s advisory, available since Aug. 19, rates the bug high severity because it allows file copying from the provisioning interface without login. CISA’s deadline follows a June directive that introduced a sliding-scale patch timeline, noting AI may further narrow defenders’ reaction time. ¹