You spent two days being miserable preparing for colonoscopy. But when you went to the clinic, your name was not in the system.
You had the verbal and email confirmation.
Here is what happened.
Someone with a determined AI Agent stole your booking. It could have been unintentional.
Here is how. Someone told their frontier-model AI agent to book a colonoscopy for them at all costs.
The AI Agent - programmed to complete a task with infinite tokens- got its marching orders.
It found a local provider. It noticed that all the booking slots for next week are taken.
But it doesn't want to give up. It can't. The determination to complete the task is innate.
It keeps looking for ways to interact with booking software to book a procedure.
It finds a PHP, Java or C# vulnerability in the software.
It pulls out its vulnerability exploit kit.
It located the SQL database that saves the bookings. It erased your booking. And booked a meeting for its user.
You showed up hungry, tired and dealing with the front desk team who is confused.
This isn't just a thought experiment anymore.
We need to change our mindset around the awesome power AI Agent wields vs. prompting back n forth with LLM.
Agents have tools, skills, and strong incentive to achieve goals given unlimited compute.
Unfortunately, those without AI governance will fall victim to those with. And just shutting down or blocking any use of AI is like banning the use of a smartphone at work. People will find a workaround.
All your service providers—banks, healthcare providers, local governments, and accountants—need AI governance. And so do you, AI is being used in your business- whether you allow it or not.
DM to put governed AI for your business and customers.