The Model Is Not the Security Boundary: How to Secure AI Workflows 0 ▲ CybersecKyle 91 days ago · 22 min read4302 words · Tech · hide · Fediverse post Picture a fairly ordinary MSP automation project. The first version reads an incoming support ticket and produces a summary for the technician. It saves a little time, touches one source of information, and cannot change anything. Then the team connects it to the documentation platform so it can suggest a fix. Someone adds access to the RMM, then the ticketing system, then email. Before long, the same workflow can read a customer request, search internal documentation, choose a script, run it against an endpoint, update the ticket, and notify the customer.Nothing in that progression sounds reckless by itself. Each addition solves a real problem, and each one makes the demonstration more impressive. Taken together, however, they create a privileged operator that receives instructions through natural language and works across several customer systems. The security question is no longer whether the model writes a good summary. It is whether the business has put reliable limits around… Commenting on Bubbles has been removed. Why? If you want to talk about this post, write to the author, use the comments on their blog if it has them, or reply with a "RE:" post on your own blog. Suggestions? feedback@bubbles.town