AI Safety Warning As Competitive Strategy

The Fear Is Real. The Timing Is Suspicious.

Jacob Coksion is twenty-seven. He spent time at OpenAI before joining Anthropic, and he left last week not for a competitor but to sound an alarm. In a BBC interview, he said the people building advanced AI are “really scared” and warned that bot swarms capable of seizing control of the internet could be a realistic threat within six months to a year. Dario Amodei, Anthropic’s CEO, published an essay the day before calling for independent oversight and an international slowdown. Sam Altman and Elon Musk agreed.

The warnings are genuine. They are also strategically timed.

Anthropic is preparing what would be a record-breaking IPO. OpenAI has been widely expected to go public as well, though Altman postponed that move in part by citing safety concerns. Both companies now have an incentive to shape the regulatory environment before their competitors—especially Chinese ones—catch up. The safety argument and the business argument are not identical. But they overlap enough to make the timing difficult to ignore.

The American Pause Meets the Chinese Sprint

While Anthropic, OpenAI, and xAI issue joint statements calling for caution, the Chinese AI ecosystem is accelerating. DeepSeek stunned the industry in January by releasing a reasoning model at a fraction of the cost its Western rivals spent, using significantly fewer Nvidia GPUs. Baidu, Alibaba, and Tencent continue shipping new models with minimal public concern about existential risk. The Chinese government has framed AI as a matter of national survival, not philosophical debate.

This is not a new asymmetry. It is the central feature of the current moment. If the United States voluntarily constrains its pace while China does not, the result is not safety. It is strategic surrender.

Jensen Huang appears to understand this. Sources familiar with a recent Goldman Sachs conference told the BBC that Huang dismissed Coksion’s claims as “baseless.” Whether Huang is protecting Nvidia’s revenue stream or genuinely believes the risk is overstated, his position is clear: the United States must maintain its lead, and pausing is a luxury it cannot afford.

Donald Trump reinforced that logic publicly. “The one who wins at AI wins,” he said, arguing the U.S. must stay ahead of China.

So the debate fractures along a line that has almost nothing to do with whether the science is sound. It runs between those who believe the risk justifies slowing down and those who believe that slowing down is itself the greater risk.

Regulatory Capture or Genuine Caution?

Critics argue that Anthropic’s safety advocacy is an exercise in regulatory capture. The company already possesses some of the most advanced alignment research in the industry. Its advisory board includes the former British prime minister, Rishi Sunak. It helped draft the earliest frameworks for evaluating AI risk. If new regulations require rigorous testing, transparency, and oversight before deployment, who is best positioned to comply? Not a Chinese startup operating in Shenzhen. Not even OpenAI, which is moving faster on commercial products.

Clément Delang, CEO of Hugging Face, put it bluntly on social media: asking Coksion about existential risk is like asking an HVAC technician about climate change. The concern may be honest, but the perspective is narrow.

George Allison, CEO of Grindr, went further, suggesting that safety warnings are partly about investor psychology. These companies need to justify valuations in the tens of billions. The narrative that their AI will dominate every industry justifies the price. The parallel narrative that it could destroy civilization justifies the regulation that keeps competitors out.

None of this proves the risk is fabricated. Geoffrey Hinton, the “father of AI” and a Nobel laureate, told the BBC that a ten percent chance of human extinction within a decade is “not unreasonable.” Evan Hubinger, Anthropic’s own head of safety research, has said privately he believes the risk of total human extinction from AI within ten years exceeds ten percent. These are not people with anything to gain from silence.

But it does prove that the people making the loudest safety claims are also the people who stand to gain the most from regulation.

What English-Language Readers Miss

The Western media treats this as a philosophical debate—a civilizational reckoning played out in op-eds and podcast appearances. That framing is incomplete. The real story is geopolitical.

China does not share the Western obsession with existential risk. Its AI strategy is instrumental, not philosophical. The Chinese state sees AI as infrastructure, like railways or semiconductors. It will fund it, direct it, and deploy it. It will not pause for a safety review sponsored by a company that is about to go public.

If the United States adopts meaningful regulation based on worst-case scenarios—even plausible ones—while China continues at full speed, the gap widens. The companies that built the safety case become the gatekeepers. Everyone else, including foreign competitors, must pass through their framework. That is not a hypothetical. It is exactly how standards and regulations have worked in every major technology shift since the telephone.

The danger is not that safety advocates are lying. The danger is that their urgency is being treated as apolitical when it is inherently political.

Who Wins. Who Loses. What Happens Next.

Anthropic wins if regulation raises barriers to entry and its alignment research becomes the compliance standard. OpenAI wins if it can delay competitors long enough to establish its products as the default. Investors in both companies win if the narrative justifies their valuations.

Chinese AI firms lose if the West successfully exports its regulatory framework. American startups without alignment infrastructure lose regardless. Consumers may lose access to cheaper, faster models if the pace of deployment slows. Everyone else lives with the consequences of whichever side moves faster.

The IPO clock is ticking. Anthropic’s last known valuation was $85.2 billion. OpenAI may still go public next year. Neither company can afford to let the narrative shift toward unchecked acceleration. The safety pause, as proposed by Amodei, gives them exactly the breathing room they need.

Whether that pause protects humanity or protects market share is the question no press release will answer.