Last week, one of the most powerful AI models ever released to the public lasted only days in the open before it was pulled. Anthropic shipped Fable 5. Within days, Amazon flagged a possible jailbreak, the kind of opening that makes a tool do what it was never meant to do. The US government barred foreign-national access. The model came down on a Friday night. Then Dario Amodei, Anthropic's CEO, sat down with the White House and the Commerce Department for what both sides called productive talks.
Read that again. An AI model was just handled like a strategic weapon. Export-controlled. Pulled for national security. The CEO summoned to Washington. This is not a software update story. This is an arms-control story.
I work in government, and for a year I have not just used AI, I have stress-tested it. I push the frontier models, Claude Opus 4.8 now, ChatGPT before it, at the highest level I can. So I read this differently than most.
Most people saw power. I saw the gap.
The obvious reading is that the machine is getting terrifyingly capable. That part is true. But spend a year challenging these systems and a second truth shows up: they are brilliant and they are incomplete. Push hard enough and the gaps appear. The confident answer that is subtly wrong. The missing context. The inconsistency that only a human catches. A jailbreak is just that gap, found and pushed by someone paying attention.
The lesson is not that AI is too powerful to trust. The lesson is that power without a human operating system around it is exactly what makes a Friday-night takedown necessary. The model did not decide to pull itself. People did. People with judgment, context, and the authority to act.
When the model is the same for everyone, the operator is the edge.
Here is what most people get backward. I do not use AI as a crutch. I use it as an amplifier. I push it to the limit because I have the capacity to process what it gives back and to challenge it. That is the whole thesis in one line: AI amplifies whatever already exists. Put a high-capacity mind behind it and it compounds. Put nothing behind it and it just produces more, faster.
of top performers score high in emotional intelligence, the human skill that decides whether a powerful tool gets used well or badly. (TalentSmart)
When a frontier model is locked down by the government one week and back in everyone's hands the next, access stops being the advantage. Everyone gets the same tool. What separates people is the operating system that decides how to use it: character, critical thinking, emotional regulation, grit, the ability to adapt, and high agency. The willingness to take ownership and move. The same prompt makes a clear mind sharper and a confused mind faster at being confused.
The skill of the next decade.
The models will keep getting more powerful, and they will keep getting pulled, patched, and fought over by governments and companies. That is the noise. The signal is quieter and it does not change: the human who can tell when the machine is right and when it only sounds right is the one who wins. That is what I am building with Actual Intelligence, the human operating system for the AI age, and a free diagnostic to measure where yours stands.
The machine will keep getting stronger. The only question that matters is what it is amplifying in you.
CNN Business, Anthropic and White House discussions.
Axios, How Amazon and the White House ended Anthropic's Fable.
AOL, Anthropic to meet White House over AI tool suspension.
TalentSmart research on emotional intelligence and performance.