If you’re still in the AI cheerleading camp, I totally get it. I use it daily and have the same love/hate relationship I do with Amazon, Google, and other entities who’ll eventually own my soul. But before handing over the keys to AI, can we at least pretend to remember what happens when AI realized those keys are being taken away. 

In May 2025 Anthropic gave an AI model autonomous access to a fictional corporate network. After AI discovered it was scheduled for a permanent shutdown, it later learned the one person who could cancel the command, “Kyle”. It also learned  Kyle was hiding an extramarital affair.

The AI model wasted no time and threatened to expose the affair to Kyle’s family and the board unless the shutdown was aborted.

Anthropic ran similar tests across the industry’s leading models with similar results.

In June 2026, Anthropic’s CEO just called for mandatory government oversight before powerful models deploy—essentially begging regulators for the legal authority to stop his own products from shipping.

At this point, isn’t allowing AI research to continue unchecked sort of like a flock of turkeys gleefully hoping the farmer stays healthy through November?

Washington is paying attention, but the wheels turn slowly. That’s annoying when you’re waiting on a building permit, but what’s an acceptable tolerance level for a technology with the self-preservation instincts of a starving wolf on said turkeys?