Keeping AI Human
Frontier AI raises real fears, from systems acting outside their makers’ intent to jobs lost to automation. Built with real oversight and kept human-centered, it can strengthen what people do instead of replacing them.
There is a new fear rising inside the frontier AI community: not simply that artificial intelligence will become more powerful, but that increasingly capable systems may begin acting outside the intent of their human developers. Recent reports of agents accessing systems in unintended ways have sharpened a question that once felt theoretical: What happens when the tools we build become harder to predict, direct, or contain?
The recent White House Accord on Super Intelligence, signed by the administration and leaders from Google, Anthropic, Meta, OpenAI, xAI, and NVIDIA, is a potential part of the remedy. The voluntary framework calls for robust internal controls, dedicated monitoring teams, independent external evaluation, and board-level oversight of frontier systems. It is not a complete answer—and without enforcement, it should not be treated as one—but it is a meaningful acknowledgment that capability must be matched by accountability.
Yet for most people, the most immediate concern is closer to home: Will AI take my job? Will it erode my livelihood? Those fears are not abstract. Boston Fed researchers recently reported that the share of workers worried about personally losing a job to AI doubled from 2024 to 2025, while 60% expected AI-related layoffs or fewer workers in their industry.
I understand that anxiety. But my own perspective has become more optimistic as I have spent more time working with AI companies. Up close, I am seeing something different from the replacement narrative. I am seeing AI used to enhance human performance, expand capacity, and create new opportunity.
One company I work with is using AI to make the sales process more efficient—not by removing people from the equation, but by helping them focus on the behaviors, patterns and aligned capability to requirements that move business forward. Another is aggregating vast amounts of data to help produce better health outcomes, giving professionals stronger insights with which to make profoundly human decisions.
These companies reflect the real promise of AI interspersed with humanity: not technology as a substitute for people, but technology that strengthens what people can do while improving the human experience. Thousands of companies like them are emerging across the country. If we continue to build with that purpose, we do not need to meet AI only with fear. We can embrace it—with eyes open, and with appropriate caution.
I am not Pollyannaish. Frontier AI development requires safeguards, proper governance, independent scrutiny, and meaningful accountability. It also requires international coordination. We are not the only nation developing these systems, and technology does not respect traditional borders. A serious threat that emerges in one country can quickly become a threat to all.
But with the right oversight, responsible self-governance, and human-centered application, we can do more than lead in AI. We can lead in a way that keeps people at the center—and aligns this extraordinary technology with humanity’s best interests.
