The Commons Letter · No. 2
Who tells the story of AI?
Dear all,
We started the Technology Commons Alliance with the mission to keep the progress of science and engineering a public good. In our second letter, we discuss who is telling the story of AI, and who is able to check it.
In August 1939, Leo Szilárd drafted a letter, signed by Albert Einstein, and had it delivered to President Franklin Roosevelt. A chain reaction in uranium, it warned, could make a bomb of enormous power, and Germany had the means to build one.
Roosevelt appointed a committee to examine the claim, and over the next two years the government drew the people who understood the physics out of their universities and into public service, where their judgment could be tested and acted on. Only then did it commit to the largest research project in American history, the Manhattan Project. The warning came from people with nothing to sell, and the government built the capacity to check it.
Today, the shape of the warning is familiar: a few people understand a technology that could change the world, and are sounding alarm bells about it. Framing AI as an existential danger, or as a radical leap forward, cements it as a key story of our time and builds public willingness to pour resources into it: money, attention and participation. Then as now, the story told about a technology is as much a lever as the technology itself.
Who tells the story today
There are a few key differences between the nuclear program of the 1940s and AI today. In 1939, the people who understood the technology were not selling it, and they communicated with the government discreetly. Today, the warning comes from the companies building frontier AI systems, and is announced publicly.
To the extent that there is communication with regulators, evaluative capacity is likely lower than it has ever been. Since the 1970s, the federal government has stepped back from technology development as the private sector expanded, and private money now finances and manages most AI development. That shift drained expertise from the state. Many scientists now work in a private job market, and the people best placed to assess these systems are usually employed by the companies that build them. For example, the US Blueprint for an AI Bill of Rights relies on voluntary commitments rather than binding legal obligations, and while the EU AI Act does impose enforceable requirements, it still permits providers to assess certain high-risk systems through internal conformity procedures.
Many people inside the industry favor more regulation. But without the means to evaluate closed systems, companies still get to tell the story of their own product. Until someone outside can check, every claim about safety, benefit or risk stays unverified, and the imagination is built by the people with the most to gain from it.
Rebuilding the capacity to judge
The answer is not simply more government involvement: we need people with real technical depth whose judgment carries weight, and who do not work for anyone with an IPO in the race. That could take many forms: public interest technologists in civil society, regulators that pay well enough to keep their talent, international evaluation bodies, universities with sustained and independent funding. Whatever the form, independent evaluative capacity has to exist as a counterweight to concentration.
The people building frontier AI inherited decades of open research, public funding and shared tools that no one could have produced alone. Whoever decides today holds that in trust, and the first test of stewardship is whether they are willing to have their work judged by someone they do not control. Distributing that judgment would also mitigate market bubbles, media hype and lack of clarity that mark AI development today.
Public imagination drives attitudes and decisions. The question we can act on now is who gets to tell the story. At Technology Commons Alliance we want to build a community that can be part of this: a space outside the labs and open to all, where claims can be tested by people whose careers, funding and valuations do not depend on the answer.
Worth reading
- Minimum Conditions for Embedding Evaluators · AI Evaluator Forum
- Better Informing the Public About Loss-of-Control Risks · METR
- We Need an Independent AI Evaluation Clearinghouse · Federation of American Scientists
From us
Yours,
The TCA team
The Commons Letter is an occasional email from the Technology Commons Alliance. Subscribe to get the next one.