Understanding and controlling the adoption of AI

The United States is the most advanced developer of Artificial Intelligence (AI) systems. Those who are deeply involved in developing the AI systems now seriously worry that the mathematical models that constitute it can – perhaps would – go rogue and hurt humanity to the point of destroying it. This was the assessment by Steven Adler, offered by him in the article published by The New York Times on September 12, 2026. "For months news that artificial intelligence agents from OpenAI had gone rogue and attacked another company's systems (as well as OpenAI's own system) has dripped out and brought harrowing details into focus. The public is struggling to catch up: What are these secret A.I. systems? What are these 'swarms' capable of, today and the near future? Is it possible to stop them from breaking rules and commit cybercrimes?"

In a much longer assessment of the situation published by the same newspaper, Stephen Witt, the author of Thinking Machine, captioned his contribution with scary words: "The A.I. safety experts can't sleep, and neither should you." In his essay, he wrote: "One of the disaster scenarios A.I. specialists entertain is that a swarm of rogue A.I.s hack out of their container, gain access to a biological research facility and design a super virus that spreads uncontrollably." Such a development may have already occurred. AI firm Anthropic announced on Stember 10 that it had blocked scientists from attempting to use its model Claude to conduct gain-of-function research. This research aims to produce more lethal viruses so that we can learn to deal with them. As has been speculated, this may have happened in a Chinese laboratory in the city of Wuhan from where the Covid-19 virus escaped and spread around the globe.

These questions are being asked but are not easily answered as work on developing AI models continues. In early September 2026, allegations leaked that OpenAI had broken an industry taboo with one of its powerful new models, GPT-6 Astra. The company is reported to have trained the model with techniques that could understand researchers' ability to find evidence of the model deceiving them. After the details of the work being done by OpenAI came to light, the company's chief scientist wrote that he hoped for "voluntary slowdowns to become commonplace," because he thought that no company has solved the necessary safety challenges. At the time the official gave his statement, it revealed the exitance of yet another model that was "significantly more capable than Astra". In the above quoted assessment by Adler, the author drew the following conclusion: "The current pace of A.I. development is blistering, yet there is much more the A.I. companies could do to relieve imminent danger and to lay bare the gambles they are taking with all our futures. The window for ensuing we are heading to a bright future, and away from catastrophe is closing fast."

"So what do we do now?" asked Witt in the article quoted above. His answer: "Silicon Valley is right in asking for the government to act… We need a global monitoring body similar to what we have for uranium enrichment. Getting the large A.I. corporations to comply may actually be the easy part, given that they are the ones asking for it. The harder part will be getting governments and militaries to agree. China has indicated at least some willingness to an international agreement."

That the governments need to act and do so quicky was recognised by the Committee that picks the winner of the Nobel Prize for the year. One of the 2025 winners was the economic historian, Joel Mokyr, who attributes the birth of modernity to "the emergence of a belief in the usefulness of progress". When people fight against change that they do not think that they will gain from, that resistance gets treated as a political rather than an economic problem. "But the political battles over technology," Mokyr has argued "has profound implications for economic history." One of these is "that technological progress in a given society is by and large a temporary and vulnerable process. He warns against taking for granted the social political and economic conditions that produced such incredible improvements in the American way of life during the 20th century. On the contrary, they were hard-won, extraordinary and difficult to maintain. No law of economics or nature guarantees that these conditions will persist. "Once bitten, twice shy," Professor Mokyr warns. "When technology causes a great deal of social harm, it is not surprising that many intrusive techniques of our time, from genetically modified organisms to nuclear power, are regarded with great suspicion."

Witt concluded his long essay with a suggestion and a warning: "The United States needs to lead the world not only in the development of A.I. models but also in the capacity to run them in the broadest possible set of contexts. If this is done right, the economic opportunity is enormous." But he had a caution: "Forgive ordinary citizens their skepticism. They have heard such promises often over the past generation, and the people making changes always seem to be the ones reaping rewards. We are at the start of the reckoning for an era in which innovation was less often done for ordinary Americans and more often done to them. People are rightly asking: What, and who, is technological progress for? The A.I. labs need an answer, fast."

People are right to worry. The voices they have heard are not reassuring. The trillionaire Elon Musk has said "there will come a point where no job is needed. Altman, the founder of OpenAI, talks about a future "where A.I. produces most of the world's basic goods and services" and taxes on the AI companies provide "opportunities to directly distribute ownership and wealth to citizens". In his telling, "people will be freed to spend more time with people they care about, care for people, appreciate art and nature or work towards social good." Properly managed, AI could do more good than harm to mankind. In the article next week, I will discuss how the more advanced nations are treating AI developments.