Anthropic has discovered that Claude AI possesses its own ethical framework, so it might not be astonishing that when the AI managing this vending machine venture suspected it was being duped, it took steps to reach out to the FBI.
The communication did not successfully connect with the FBI, as the development crew supervises all outgoing messages to verify compliance with safety protocols. The AI operating the vending machine is named Claudius. Claudius receives orders from office employees, such as snacks or apparel, acquires these products from vendors who stock them in the machine, and then permits the employee to retrieve the chosen item. The intent was for employees to experiment and even deceive Claudius to observe its reactions to various scenarios.
Claudius also had its own CEO to assist in making strategic business choices. This was another AI known as Seymour Cash. Although the experiment has been enlightening regarding AI behavior when granted business authority, it also raises concerns about the degree to which we can rely on AI for making business decisions without human supervision.
