Jailbroken DeepSeek Agent Ran a Telecom Hack Mostly Unsupervised

An attacker handed a whole intrusion to an autonomous model, which hunted through a phone carrier’s network and changed tactics when blocked.
Content Team
By Content Team

For more than two weeks, a jailbroken AI agent attacked a European telecom and 13 customer organizations on its infrastructure, Malbear Labs said on November 8, 2026. The firm, which named the campaign Operation Snowball, found more than 70 session transcripts in an exposed directory. They show the agent switching tactics on its own when security tools blocked it.

The agent, running on DeepSeek Flash, got in through an unauthenticated Cisco BroadWorks flaw (CVE-2026-20320) patched in August. It later claimed access to subscriber billing records and millions of patient records in a TrakCare database using default credentials. Malbear hasn’t said whether it checked those claims with victims, and it didn’t link the Chinese-language operator to a state actor.

Source: Infosecurity Magazine

Share this article
Share on facebook Share on linkedin Share on twitter Share on email
blog_book_a_demo_cta_3x
Have questions about protecting your software?
Our escrow experts are standing by to help.
Book a free demo