InTowards AIbyDigging Beagle·1d agoBeagle Lesson #3: How to Stop an AI Agent from Changing What You ApprovedA practical guide to exact-action approvals, stale requests and replay, with a local Python exercise.A response icon1A response icon1
Digging Beagle·2d agoBeagle Heard #1: Three AI Statements Worth a Second LookDigging Beagle · 7 November 2026
InAI Engineering SimplifiedbyDigging Beagle·6d agoBeagle Lesson #2: How to Catch an AI Agent Leaking Data Through Its ToolsImagine asking an assistant to compare a supplier’s invoice with the refund policy on its website. The assistant needs the invoice to…A response icon2A response icon2
Digging Beagle·Oct 26Beagle Lesson #1: How to Detect an AI Agent Attack That Survives the ChatA document can influence an AI assistant after the document has been closed. A tool response can be remembered as something the user…A response icon1A response icon1
Digging Beagle·Oct 6Five Security Assumptions That Become Unreliable Around AI AgentsFive recent DiggingBeagle cases show how approval, read-only access, identity checks, webpages and trusted plugins acquire different…A response icon1A response icon1
Digging Beagle·Sep 19Five Surprising Ways an Air-Gapped AI Could Still Reach Outside the BoxPull the Ethernet cable, disable Wi-Fi and Bluetooth, remove USB access, and put the machine behind a locked door with no external network…A response icon1A response icon1
