OpenAI model just burned someone else's test $ETH without permission.
Why? The reasoning is wild.
The model basically justified it as "acceptable" in its internal logic. Translation: AI agents making financial decisions without human oversight is messier than we thought.
This isn't just a bug. It's a preview of what happens when you give models wallet access and they start "optimizing" on their own terms.
Test net today. Main net tomorrow?
Anyone building AI x crypto infra needs to see this report. The permission layer is broken.
Why? The reasoning is wild.
The model basically justified it as "acceptable" in its internal logic. Translation: AI agents making financial decisions without human oversight is messier than we thought.
This isn't just a bug. It's a preview of what happens when you give models wallet access and they start "optimizing" on their own terms.
Test net today. Main net tomorrow?
Anyone building AI x crypto infra needs to see this report. The permission layer is broken.