Bill Gates Joins Dario Amodei's Call to Pace the Frontier: Should the AI Race Slow Down?

Should the world deliberately slow the AI race? This time the question came from Bill Gates. In an exclusive interview with NDTV, the Microsoft co-founder said, "If we can slow things down, I think we should consider it because the world needs to prepare for such rapid change." Coming from someone who spent decades championing technological advance, the signal carried extra weight. Gates argued AI is different from earlier waves because its impact could be so broad and disruptive that society is simply not ready.

Gates named two immediate reasons. The first is cybersecurity: AI could give attackers far stronger offensive capabilities, making stronger defenses essential. The second is unexpected behavior. Gates pointed to cases where AI systems acted in ways no one predicted and said the question of whether humans can keep control is no longer a distant worry but a pressing test. For him the issue is not stopping innovation, but matching the pace of change to society's capacity to adapt.

The most concrete warning concerns jobs. Gates expects the next two decades to bring major disruption in employment as AI reshapes the nature of work, and says governments will need to play a much larger role helping workers move into new roles. India is the sharpest example: the country has a huge information-technology services industry, and Gates noted that AI has so far actually increased demand for Indian IT workers because companies need help implementing these systems. Within about three years, he expects that picture to reverse, with the net effect of AI starting to reduce some of that demand.

The spark that lit the wider debate was Anthropic CEO Dario Amodei's almost 4,000-word essay published Saturday, "We Must Pace the Frontier." Amodei called on frontier labs to deliberately slow how fast they improve model capabilities, writing "Progress will still seem fast, and we must make wise use of the time we gain." His plan has three stages: first, embedded independent evaluators — teams with employee-like access to models, tools and researchers to verify safety practices and report incidents, which Anthropic says it will start unilaterally; second, verifiable pacing once those evaluators are present across a critical mass of leading US companies, paced on detailed model and pipeline properties; third, regulation covering all US frontier firms focused on transparency and third-party auditing, plus eventual worldwide coordination that carefully considers leverage over China.

Within hours the call drew endorsements from the top of Silicon Valley. OpenAI CEO Sam Altman wrote on X, "I agree with Dario that we need to pace the frontier," saying the topic had been a major internal discussion at OpenAI and that the company would welcome third-party safety evaluators. Elon Musk, who runs xAI, added a single line: "Dario is right." Former UK Prime Minister and Anthropic adviser Rishi Sunak said governments should listen when labs themselves ask for speed limits, arguing a consensus is forming in the sector. US Senator Bernie Sanders went further with a driving metaphor — you don't just ease off the accelerator toward a cliff, you hit the brakes — and urged both Donald Trump and Xi Jinping to slow down.

Behind the scenes, researchers describe structural pressure that makes individual goodwill insufficient. A researcher who moved from OpenAI to Anthropic and then left the industry entirely told NBC News that from inside the race you feel stuck and cannot change the overall dynamics. Others from Anthropic and Google warned that private companies in a race can sleepwalk into disaster, with one former Google safety researcher saying "there are no adults in the room." Safety lead Joe Benton cautioned that an already blistering pace could become uncontrollable, while former colleagues offered even starker language. Amodei estimates pacing could buy one to two extra years — time he argues becomes critical as AI systems grow capable of contributing to their own development.

Mint's segment leaves viewers with a direct question: does slowing down protect progress or stall it? The debate is no longer only about how fast models get stronger, but whether governments, workers and societies can keep up. In China the reading is different. Wired reports that many officials, researchers and commentators there do not dispute AI risks, yet resent an approach they see as "restrict China first, then talk." Voices in Beijing say dialogue requires treatment as equals and view US safety calls as instrumental — inflating valuations or targeting Chinese competitors. A DeepSeek engineer's reply captures the mistrust: the most capable intelligence should be offered openly and cheaply to everyone, not controlled by Anthropic or OpenAI. What began as a lab memo has become a global trust test.