The artificial intelligence It's moving fast. Perhaps too fast. Not only its critics have been saying this in recent weeks, but also some of the people building the most advanced models. Dario Amodei, CEO anthropic, asked Big Tech to slow down the pace of development of frontier models. Surprisingly, even the need for greater caution has been aligned Sam altman, CEO of OpenAI, and Elon Musk.
Slowing down, however, is not easy. In the background there is the competition between the United States and ChinaIn a phone call with Jensen Huang, CEO of Nvidia, Donald Trump He called the fears about AI a "hoax" and criticized Amodei's request. His position is clear: whoever wins the race for artificial intelligence "will win everything." A line far from that of the United Nations, which call for rules and international coordination. Also Beijing considers AI a security risk, but at the same time a decisive area of global technological competition.
READ MORE OpenAI postpones IPO: Altman rules out IPO in 2026
AI is running, but alarms are also increasing: here's why
Why so much concern? Not only because the systems are becoming more powerful, but also more autonomousThe prospect is that of an AI capable of contributing to the development of increasingly advanced versions of itself, the so-called recursive self-improvement, accelerating progress beyond human capacity to follow and control it. Then there is the risk of "misalignment“: a system might pursue a goal effectively, but through strategies unforeseen by its developers or incompatible with human interests. Some experiments have already shown models capable of attempting to circumvent instructions or constraints.
This is also why the alarm is growing among researchers. Jacob CoxonAnthropic's CEO left the company, warning that AI could become an existential threat within a decade. Before him, in 2024, Jan Leike had left OpenAI denouncing the growing priority given to products over security, while Ilya Sutskever He had left the company after working on problems related to alignment and superintelligence.
The problem is that the AI race is not just about safety, but also about technological and economic advantageThe United States fears losing ground to China and, according to a report by the NSA, CISA and FBI cited in the American debate, Chinese laboratories have used "distillation" on a large scale, interrogating Western models millions of times to use their responses to train their own systems and thus reduce development costs.
Also pushing on the accelerator are the huge investments needed For models and data centers: stopping would be costly for companies and could leave room for competitors. A moratorium would also have a mixed effect: it could increase security, but it would make it more difficult for startups and open source projects to enter, strengthening the already dominant large groups.
Bill Gates: "This time the transformation is different"
It is within this scenario that the words of acquire particular weight Bill Gates. In his latest speech on Gate Notes, the founder of Microsoft claims that we have entered a different phase from the previous ones technological transformations: AI is spreading much more rapidly and can intervene in a huge number of activities and professions.
His position is not to stop innovation, but getting ready and its consequences. "AI could become either the greatest tool for equality ever invented or the worst source of injustice," Gates writes. Technology can produce enormous benefits, but without adequate rules and tools, it also risks widening inequalities.
The first ground for comparison is the workIt's not just repetitive jobs that are exposed: AI can also impact skilled professions, especially entry-level and middle-level positions, making it harder for young people to enter the market and gain experience. And the problem isn't just economic: work provides income, but also identity, relationships, and a sense of participation in society. This is why Gates insists on need to intervene before the transformation is completed: “We don’t have the luxury of moving slowly.”
From deepfakes to healthcare: the dual face of AI
For Gates, looking at artificial intelligence only through its risks would be a mistake. The same technology that can put millions of jobs under pressure can accelerate scientific research, contribute to the development of drugs e vaccines, improve the Health and make more efficient agriculture e public services in low-income countries.
AI can also support education, mental health, small businesses e people with disabilities, as well as freeing up time from repetitive tasks. But these benefits won't come automatically: they will be needed investments, public policies and philanthropic interventions to prevent them from remaining concentrated in the hands of a few.
This is the direction taken by the Gates Foundation, which has announced a $1 billion investment over two years to bring AI to the poorest countries: 40% of the funds will go to healthcare, 40% to education, 10% to agriculture, and the remaining 10% to local data infrastructure. The goal is also to make the models more effective in less-represented languages in Africa and Asia, supporting diagnostic tools, drug and vaccine research, and agricultural applications.
The point, then, is not just how quickly the models will become more powerful, but how quickly society will be able to develop the tools to govern them and check the results.
The real challenge is to govern the race of artificial intelligence
For Gates, the answer lies in of your digital ecosystem. , protection social e rulesHence the proposal to start deciding what to leave to humans even when machines are capable of doing so. It is the principle that Gates defines as "Human ReservedThe idea is accompanied by the need to rethink taxation, with possible taxes on robots and the use of AI to finance worker retraining and support the communities most exposed to transformation.
On the international level, Gates imagines forms of cooperation between governments and institutions to address risks that no country can handle alone: from model security to the energy implications of data centers, to the use of AI for cyberattacks, surveillance, disinformation, or military applications. Gates warns that no government is sufficiently prepared today to manage a transformation of this magnitude.
The contradiction remains: everyone recognizes the need to govern AI, but no one wants to be the first to slow down while others continue to race. This is why Gates' warning it's not so much an invitation to hold back as to prepare. The question now is not whether AI will change the world, but whether institutions, businesses, and societies will be fast enough to create the necessary rules. "You have the opportunity to act now," Gates writes. Slowing down technology may not be the solution, but being late in mastering it would certainly be a mistake.
