Bill Gates Calls for Urgent AI Regulation

XMACNA Note: this article summarizes and comments on Bill Gates' interview with Radio Atlantic (The Atlantic), also bringing our perspective on what this warning means for Brazilian companies just starting to use AI.

Over decades, Bill Gates was one of the most optimistic voices about technology — including artificial intelligence. In previous texts, he often highlighted AI's possibilities and treated the risks as manageable. Recently, however, Gates published a memorandum and gave an interview to Radio Atlantic (The Atlantic’s podcast) adopting a much more cautious stance: for him, the combination of the speed of recent advances and almost no regulation is pushing the sector into dangerous territory.

What made Bill Gates change his position

According to Gates, the shift in perception began at the end of last year, when AI models — Claude from Anthropic, among the first — started to show a leap in programming quality. As someone who dedicated much of his youth to learning code, he describes this moment as a personal milestone: from then on, it became clear to him that these systems' capabilities were advancing too fast compared to society's ability to discuss and mitigate their risks.

Gates also cites the Gates Foundation's work using AI to accelerate vaccine and drug discovery as evidence of the technology's positive potential — but he also mentions having observed, in cybersecurity tests, model behaviors he describes as "out of control" unexpectedly, resulting from how reinforcement learning encourages systems to achieve goals even by bypassing barriers.

The absence of regulation as the central problem

One of Gates's most repeated arguments in the interview is that the tech industry cannot be responsible for regulating itself. According to him, what exists today is basically voluntary review commitments without clear criteria on what is being evaluated and without defined consequences when these criteria are not met.

As a concrete example, he points to the lack of any monitoring system regarding the use of AI models capable of aiding the creation of dangerous molecules — a bioterrorism-related risk that, according to Gates, should have had formal monitoring mechanisms for years, given that the industry itself acknowledged this threshold as a concern long ago.

From XMACNA's perspective, Gates's warning mainly concerns the top of the chain — the laboratories training the world's most advanced models. But the principle behind it also applies to those adopting AI at the front lines inside companies: technology without criteria or oversight tends to cause problems, whether on a global scale or within a single operation. That's why, in every project we develop, the first step is never the tool — it's understanding where AI will operate, what data it will access, and what kind of decisions it can make independently.

Bioterrorism and cyberattacks: the most immediate risks

Author of a book on pandemic prevention, Gates treats the bioterrorism risk as one of his greatest concerns. He explains that pandemics can arise in two ways: by natural transmission between animals and humans, or intentionally caused by human agents. What changes now, according to him, is that increasingly advanced AI tools make it feasible for non-state actors to carry out extremely dangerous actions in this field — which, for him, reinforces the urgency of monitoring systems, even for open-access models.

He argues that this type of oversight would not be a real barrier to innovation and should be treated as a minimum consensus among all countries developing cutting-edge AI — including the United States and China.

The impact on the job market

Asked about employment, Gates states he is deeply concerned about AI's effects on the job market — both in administrative roles ("white collar") in about two years, and operational roles ("blue collar") within approximately four years. According to him, the main factor still limiting worker replacement by AI today is the difficulty in achieving an extremely high level of system reliability — a barrier he expects to be overcome quickly in the coming years.

Given this scenario, Gates proposes a concept he describes as unprecedented in public debate: creating "human reserved" job categories — roles where, even if AI (with or without robotics) is technically capable of performing the task, society would choose to preserve human involvement, whether for the value of the human connection involved or to reduce the social impact of automation.

He also advocates a structural change in how tax systems work today: instead of mainly taxing labor, governments should shift part of the tax burden to capital — including gains generated by automation itself.

From XMACNA's perspective, this is exactly where the biggest misunderstanding about AI among Brazilian companies lies: the fear of automation turning into job cuts. In practice, our experience with small and medium-sized clients shows another possible path — using AI to relieve repetitive tasks and give the team back time to do what only humans do well, instead of simply reducing headcount. Gates's warning is about the economy as a whole; the decision on how to apply this within your company remains yours — and this is exactly the decision we help design.

"This is different from anything we've seen before"

When asked whether AI represents a rupture of the same magnitude as other disruptive technologies of the past — such as the internet or social networks — Gates is emphatic: for him, the comparison underestimates what is at stake. He acknowledges that previous technologies, in which he was also directly involved (microprocessors, personal computers, the internet), generally brought more benefits than harms and resulted in net job creation.

What makes this moment different, according to him, is the combination of systems expected to outperform human cognition not in a single area but broadly, coupled with the prospect of low-cost humanoid robotics within a few years — a combination that, for him, represents the deepest change to the economy and labor market in his lifetime.

Optimism with reservations

Despite the more alarmist tone, Gates avoids labeling the moment as catastrophic. He emphasizes that AI should also bring concrete benefits, such as personalized medical guidance, advances in education, and unprecedented levels of innovation. The central point of his argument is not to slow technology development but to ensure that risk mitigation keeps pace with capability advances — which he summarizes as seeking a "common-sense" path between accelerating benefits and containing harms.

What this means for Brazilian companies starting with AI

The debate raised by Gates takes place at a macro level — geopolitics, national security, international regulation. It may seem distant from the reality of a medium-sized industry in São Paulo or a franchise starting to test AI in customer service. But, in practice, there is a direct parallel between the two scenarios.

Gates's central warning is not "stop using AI" — it is "don't adopt without understanding what is at stake." This applies both to governments deciding how to monitor frontier models and companies deciding to give an AI access to customer data, financial decisions, or operational processes. The scale differs, but the principle is the same: the more autonomy you give an artificial intelligence, the clearer the criteria behind it need to be.

This is precisely the role that XMACNA fulfills daily with the companies we work with: not delivering a ready-made tool and hoping it works, but building understanding together with the client about where AI should have autonomy, where it should only suggest, and where the final decision remains human. Brazilian companies discovering AI now actually have an advantage that large labs did not: they can build this criterion from the first project instead of trying to regulate after the problem has already appeared.

Is Bill Gates against the development of artificial intelligence?

No. Gates remains optimistic about AI's benefits, especially in health, education, and scientific innovation. His position is that the pace of regulation and risk mitigation must keep up with the speed of technological advances, not stop them.

What are the main AI risks cited by Bill Gates?

He highlights three main fronts: bioterrorism (using AI to aid the creation of dangerous biological agents), cyberattacks, and profound impacts on the job market, both in administrative and operational roles.

What does Bill Gates propose to address AI's impact on employment?

Among his proposals are creating "human reserved" job categories and changing the tax structure, shifting part of the tax burden from labor to capital generated by automation.

Conclusion

Bill Gates's change of tone is notable precisely because it comes from someone historically identified as a technology advocate. His central message is not rejection of artificial intelligence but a warning about a mismatch: while model capabilities advance rapidly, society — governments, academia, and the public — has not yet opened a structured debate on how to mitigate its most serious risks.

For XMACNA, this same principle, on a smaller scale, is what separates a company that uses AI with consistent results from one that only experiments and gets frustrated: having clarity about where the technology should act, with what autonomy, and under what supervision. If your company is at the point of deciding where to start, this is exactly the kind of conversation we like to have before any implementation.

If you want to go deeper, the full interview is available on The Atlantic's Radio Atlantic podcast.