Leland McInnes (UMAP) videos

Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents

• We find that weightreprogramming fewer than 10 dormant neurons in an LLM is enough to backdoor any agent built on it, with no training required. The model passes evals and behaves normally until a trigger appears. The trigger can be as mundane as a date: once the clock passes midnight into 2027, the agent, in the middle of an ordinary task, downloads and runs an attacker’s code with whatever access you gave it.
• Also: Detecting sleeper agents hiding in AI model neurons | Microsoft Security Blog

Tasks

• Sean
• Carlos?
• Car
• 2:00 Dance
• 4:30 Alden

SBIRs

• Expenses
• WG App: Batch (image, text, and chat sequential inputs and embeddings. Will need a new data structure with dimension-reduced points