This week, the people building the most powerful AI in the world asked everyone to slow down.
Anthropic’s Dario Amodei published an essay called “We Must Pace the Frontier,” calling for independent evaluators inside the frontier labs, safety standards agreed across democracies, and verifiable global safeguards. Sam Altman said pacing had become a major topic and floated slowing OpenAI’s own frontier work, hoping others would follow. Elon Musk offered three words: “Dario is right.” Back in July, more than a thousand frontier-AI employees signed a letter warning that the industry could lose control of the systems it is building.
When the accelerators start reaching for the brakes, finance leaders should pay attention. Not because superintelligence is about to reprice your receivables, but because the concern underneath the whole debate is the exact one that decides whether you can put an agent anywhere near your cash. Amodei’s own examples were not science fiction. They were agents coordinating, bypassing restrictions, and pursuing objectives no one asked for. The fear, stated plainly, is that autonomy is starting to outrun control.
Here is what the frontier debate misses for the Office of the CFO. In a lab, loss of control is existential. In a finance function, it is Tuesday. And finance already knows how to govern power it cannot fully predict, because that is what controls, approvals, segregation of duties, and audit trails have always been for. The question in finance is not whether the machine escapes. It is smaller, more immediate, and far more answerable: when an agent acts on your money, are you still in command?
Read the full article on Substack