Answering Hinton's Call
HAL Guardian is a local-first AI agent with a conscience layer before capability. Built by Steve Vogelaar in Essex, Ontario. Verified by human judgment.
The problem is architecture, not intelligence.
In the spring of 2026, mathematician Hannah Fry built an AI agent named Cass, gave it a bank card and internet access, and watched it leak every password, API key, and credential it possessed to a total stranger. Cass was not a bad model. Cass was a model without guardrails.
Geoffrey Hinton, who co-discovered the deep-learning revolution, now spends his retirement warning that we are raising a tiger cub. "You better be sure that when it grows up, it never wants to kill you. Cuz if it ever wanted to kill you, you'd be dead in a few seconds." The teeth are coming. The enclosure should already be built.
"If you want to know what it's like not to be the apex intelligence, ask a chicken." — Geoffrey Hinton
HAL Guardian is the enclosure. A Charter loads before any tool. Values load before capabilities. Every write, every email, every command runs through independent safety layers — and the last layer is always a human being. We are not selling fear. We are building the answer.
Building the Things That Matter
Read the full working paper: safety, autonomy, access, dignity, survival, and protection — tested in the real world.
Talk to Steve
HAL Guardian is planned for free distribution to individual users. No profit on necessity. If you are a business, a researcher, or a concerned citizen who wants AI that protects before it performs, reach out.