Where do you draw the line on handing autonomy over to AI?

A neuroscientist, a cardiologist, an oncologist, and a radiologist walk into a Toronto bar and order mocktails. All four of us are entrepreneurs. None of us wears just one hat.

It sounds like the setup to a joke, but it quickly turned into a fascinating conversation on AI and trust, one of the best I have had all year, at our Toronto Pamir Ventures dinner.

The question I asked friends this week was simple: where do you draw the line on handing autonomy over to AI?

When you put clinicians and founders in the same room, the conversation moves past the theoretical hype very quickly. We spent time dissecting what the frontier labs are shipping versus what we actually trust in practice.

We debated the real boundaries:

• Which would you trust (Muze, Instinct, Google, or OpenAI) with autonomous access to your inbox and credit card?
• Where does clinical decision support end and physician liability begin?
• When do you trust a model over human clinical judgment in life-or-death diagnostics?
• How will growing up with synthetic copilots fundamentally alter the next generation of kids?

For medicine, the consensus was clear: algorithms can accelerate workflow and surface signals, but human accountability cannot be outsourced. When outcomes matter, trust is earned through ground truth and clinical validation, not benchmark scores.

For our personal lives, the lines were even more revealing. Across the world, my partners Roger W. Ferguson, Jr. and Anand Subramanian were having similar conversations.

Surround yourself with people who constantly push the envelope, challenge your assumptions, and do not fit neatly into a single box.

Where is your personal line in the sand when it comes to trusting AI?

Originally posted on LinkedIn.