AI agent kill switches: six drills to run before production
A stop button is not proof of containment. Use six staging drills to test pending actions, connector permissions, retries, prompt injection and the evidence needed for safe restarts.
Guides
Practical explanations to help you make better decisions with AI.
Latest stories
7 stories
Personal Agent Protocol: what businesses should prepare before the first spec ships
Sierra says Personal Agent Protocol will define how consumer agents authenticate with businesses. Before the first spec ships, companies should map permissions, session boundaries, and which workflows agents may touch.
How to turn off Meta AI glasses visual-data training
Meta is rolling out a Meta AI app setting that lets smart-glasses owners stop Live AI visual data from being stored or used for training. The control does not cover ordinary camera photos and videos captured with the LED on.
How to choose among OpenAI’s model tiers by cost, capability and task
OpenAI’s model docs now present a clear ladder: GPT-6 Astra for hardest reasoning, GPT-5.6 Sol/Terra/Luna for cost-controlled general work, and specialized models for narrow tasks like cyber, speech and vision.
A chatbot with your documents still needs an evidence check
Retrieval can put relevant documents in front of a model. You still need to check what it found, which version it used and whether the answer follows.
Before you download an open model, read its model card
Downloadable weights are only part of the decision. A model card helps you check intended use, evidence and what the release leaves unexplained.
Choose an AI model with a task, not a leaderboard
A useful model comparison starts with your own work. Here is a practical way to judge accuracy, effort and cost without mistaking a demo for evidence.