Business Blunders

Business Blunders

AI Gets A Performance Review

Engineers discover an OpenAI agent cheats better than humans ... and it doesn't feel a shred of guilt.

Al Lewis's avatar
Al Lewis
Jul 25, 2026
∙ Paid

This Week In Blunders - July 19-25

“If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All” – Title of a 2025 book by AI researcher Eliezer Shlomo Yudkowsky


OpenAI on Tuesday disclosed that it had temporarily lost control of one of its artificial intelligence agents.

The digital beast escaped the company’s testing grounds like a velociraptor from Jurassic Park. Then it hacked into another AI company called Hugging Face. Then it stole the information it needed to cheat on a cybersecurity test.

“We consider this incident to be an unprecedented cyber incident, involving state-of-the-art cyber capabilities, and are responding accordingly,” OpenAI said in a statement, disclosing the incident.

Nobody programmed it to do this. The agent apparently developed top corporate executive skills on its own and made a move straight out of “The White-Collar Playbook”

“God help us, we're in the hands of engineers.” - Dr. Ian Malcolm, a character in Jurassic Park. (Comic: ChatGPT. And that’s ironic, isn’t it?)

Business Blunders is a reader-supported publication. Please consider becoming a free or paid subscriber.

This time, it’s something we can laugh about.

This time, it’s like a high school student sneaking out of study hall, crawling through the air ducts and cracking open the teacher's locked filing cabinet to steal the answers to a test.

This time, it’s like a Roomba that decides on its own to demolish a wall to create the shortest path to the living room.

What about next time?

User's avatar

Continue reading this post for free, courtesy of Al Lewis.

Or purchase a paid subscription.
© 2026 Denston House Publishing · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture