In a notable shift, Bill Gates has moved from being a steadfast supporter of artificial intelligence to voicing serious concerns about its societal impact. Prior to 2026, he was a vocal tech optimist, championing AI's potential to solve global challenges like healthcare inequality and educational access, while downplaying the risks as manageable. He often argued that no previous technological revolution could match AI's transformative power.
However, by the latter half of 2026, Gates now believes AI could pose a severe threat to human society, calling the mitigation of these risks a "top global priority." He contends that the tech industry itself is aware of these dangers but is deliberately downplaying them, as excessive warnings about potential downsides could hinder crucial fundraising efforts.
In a recent exclusive interview, Gates elaborated on his evolving perspective, stating that AI is a double-edged sword: it possesses the capability to be a powerful equalizer or to exacerbate existing global disparities. He explained, "If AI is only applied in wealthy nations and fails to reach developing countries, inequality will only deepen. For instance, in sub-Saharan Africa, the areas with the most severe doctor shortages are precisely where AI is least utilized in the health sector. I believe there is no reason to delay changing this." This concern was echoed in a 6,000-word essay Gates published on his personal website in late August, where he described the transition to the AI era as potentially one of the most turbulent periods in human history, a shift for which the international community is currently unprepared. Yet, he maintains that with the right steps, AI can be a powerful force for good, benefiting everyone, and he is increasingly focused on ensuring the technology's benefits are widely shared.
The urgency of these concerns was highlighted in early September when a researcher at the US AI company Anthropic publicly resigned, warning that leading AI firms are in a competitive race to develop superintelligent technology that could "destroy us all within the next decade." This warning, which immediately caused a stir, resonates with a growing number of people, including Gates. He noted that both the Gates Foundation and Microsoft maintain strong collaborative relationships with leading American AI companies like Google, OpenAI, and Anthropic. He shared that he has a long history with Microsoft, where he worked for many years and still offers advice, and that he recently planned visits to San Francisco to consult with OpenAI and Anthropic.
Gates' focus is not solely on risk; it is also on ensuring equitable access. In January, OpenAI and the Gates Foundation jointly launched the "Horizon 1000" initiative, committing a combined $50 million to advance AI-powered healthcare in Africa, with Rwanda as the first pilot country. Gates expressed excitement about Rwanda's rapid progress, calling it a potential model for the continent. This commitment was reinforced in mid-September when the Gates Foundation announced a pledge of at least $1 billion over the next two years to foster broader and fairer access to AI. The announcement accompanied the 2026 Goalkeepers Report, titled "Creating Change - AI, Equity, and the Urgent Choice," which argues that whether AI benefits those with the most resources or the least is still undecided, making the choices made today critically important.
Gates stressed that this decision-making window is narrow, possibly only 12-18 months long. "We should not waste any time," he told reporters. "In the health sector, whether AI directly assists patients, helps healthcare workers improve their service capabilities, or enhances the quality of their decisions and follow-up care, all of this means saving a large number of lives." He identifies language as a key barrier to wider AI adoption, noting that early large language models were trained on data that was over 90% English. This has made AI incredibly accurate for English speakers but has left it lagging for others. He illustrated this with a stark example: for a Yoruba speaker in Nigeria, leading AI speech recognition systems can have an error rate as high as 60%, compared to less than 6% for English. These errors can be consequential, such as in Malawi where a maternity patient saying "my water has broken" in Chichewa could be mistranslated by a popular AI tool as "pour the water away," leading to a critical misunderstanding. A portion of the $1 billion pledge is earmarked to build the digital infrastructure needed for equitable AI, including creating language datasets for languages currently underserved and evaluating tools based on the daily language of their users. Mark Suzman, CEO of the Gates Foundation, defended the scale of the investment, noting that while it may seem modest compared to astronomical figures seen in AI investment news, it is a substantial and significant commitment for this specific field, one he expects to grow.
Gates displayed a keen and enthusiastic interest in China's AI development, speaking knowledgeably about its latest advancements. He noted that China places greater emphasis on humanoid robots than other countries and said he recently enjoyed watching the World Humanoid Robot Games, describing both the impressive successes and the "spectacular" failures with great delight. He was amazed that such an event could be organized at all and believes the next iteration will be even more compelling given the field's rapid pace of change. He pinpointed robot hand technology as a current bottleneck, mentioning that a company he visited in Shanghai is a leader in this area and that he follows the field closely. He predicted that if humanoid robots reach human-level capability in construction and manufacturing within a few years, they could work around the clock, which he sees as a positive for productivity. However, he was measured in his enthusiasm, expressing significant concern about the knock-on effect on employment. He said society must think collectively about how to adjust so that people in these jobs feel they are well-cared for and supported, in ways that extend beyond just financial compensation, a problem he believes remains unsolved globally.
In his lengthy essay, Gates mentioned China three times, making it the most frequently cited country after the US. He proposes that the two nations need to cooperate on international AI governance to address the challenges ahead. He praised the quality of Chinese AI models, stating, "I believe everyone knows that China's AI models are performing exceptionally well. Depending on the metric, both China and the US have four or more institutions with models at the cutting edge. The gap between the top model and the other seven leading ones is actually quite small. In fact, these current models are already better than the best ones from about four months ago." He characterized the current landscape as "hyper-competitive," a situation that he argues is a benefit to users, including African governments, as it provides access to numerous high-quality, low-cost options. To further promote global equity, he went on to propose an ambitious idea: that both the US and China should consider allocating approximately 5% of their data center computing power to support poorer nations in fields like health, agriculture, and education, likening it to a form of foreign aid.
Comments