Sam Altman Indicted by ChatGPT
Busted for opportunistic China hawkism by his own large language model!
Over the past few months, in various issues of this newsletter and various episodes of its podcast, I’ve offered a cynical explanation for some of the China hawkism that has become so prevalent in Silicon Valley. Namely: Depicting China as a grave peril to America is, for big AI companies, a smart lobbying tactic. By attributing existential importance to “winning the AI race against China,” these companies can convince Washington to cough up subsidies and other forms of support and to hold off on regulations that might hobble America’s large language models as they work tirelessly to fend off the Communist peril.
Of course, this is just a hypothesis. I have no way of peering into the mind of OpenAI CEO Sam Altman (whom I’ve been particularly hard on) and finding out whether his actual concern about China matches his professed concern. However, there is a way to subject my cynical hypothesis to at least a preliminary test: If the hypothesis is on target, and much of this China alarmism is driven by lobbying goals, you’d expect there to be a correlation between how much a given company plays up the Chinese threat and how vigorously and ambitiously the company lobbies for those goals.
But, you may ask, how would we go about establishing such a correlation? How can we gauge metrics like “how much a company plays up the Chinese peril” or “how ambitiously the company lobbies”? To which I reply: Let AI worry about that! Just tell a good large language model to rank America’s big LLM makers in terms of (a) how much they play up the Chinese peril; and (b) how ambitiously they lobby for government help. Then see if the companies are ranked in the same order on the two lists. That would be a good rough and ready test for the correlation I’m predicting.
As it happens, America’s three most prominent LLM makers—OpenAI, Google, and Anthropic—recently provided data sets well suited to this purpose. The Trump administration had invited input on AI policy, and all three companies submitted written responses. So I had NZN’s crack AI special ops team feed the three responses into America’s most venerable LLM—OpenAI’s ChatGPT—and ask it to rank the companies along two dimensions: (a) “the degree to which they emphasize the need to beat China in the AI race” and (b) “the degree to which they appeal for government support either via subsidies or via loose regulation.”
If my cynical hypothesis is on target, you would expect that the same company would take the gold medal in both categories. And, indeed: the winners are… OpenAI and OpenAI!
As for the first gold medal, for China hawkism: OpenAI, its LLM noted, “explicitly anchors much of its response around the geopolitical AI race vs. the Chinese Communist Party (CCP).” This response features “multiple references to the CCP as the primary competitor, especially regarding DeepSeek (a Chinese frontier model), portraying it as a risk to freedom, privacy, and US security.”
Meanwhile, on the lobbying front, OpenAI “strongly pushes for federal preemption of state-level AI regulations” and “calls for liability protections” that permit companies to innovate “without fear of excessive legal exposure.” Plus, OpenAI wants a “large-scale government-supported buildout of energy, fiber, and compute infrastructure.” Also, if the federal government could “use tax credits, loans, sovereign wealth funds, and guarantees to catalyze private investment,” that would be great!
Would you like anything else, Mr. Altman? Why yes, actually: Could Washington please provide “offtake agreements to stabilize demand for AI compute”?
In sum, says OpenAI’s AI, “OpenAI is very proactive in seeking both regulatory relief and significant government-led infrastructure/subsidy efforts, framing them as crucial to US competitiveness.” Indeed, the company “explicitly frames compliance burdens as a national security risk (benefiting China if US companies are bogged down).”
So congratulations to Sam Altman on winning the gold medal in not one but two categories: Chinaphobia fomentation and Regulatory parasitism. Not since Jesse Owens won four golds at the 1936 Berlin Olympics has an American demonstrated such versatile talent in facing down an authoritarian adversary.
Now on to the silver medals! At this point—when ChatGPT tells us who occupies the number two slots on the two lists—my hypothesis encounters what might seem, to the naive observer, to be trouble. However, in the eyes of non-naive observers (me, for example), ChatGPT’s superficially problematic silver medal awards amount to remarkably nuanced confirmation of my characterization of the roots of Silicon Valley’s China hawkism.
The key point to appreciate before I unveil the silver medal winners is that I’ve repeatedly made a distinction between OpenAI CEO Altman and Anthropic CEO Dario Amodei: Though both emphasize the importance of winning the AI race with China, and both of their companies play the China card as part of their lobbying pitch, I’ve said that Amodei’s China hawkism seems more genuinely ideological, and less opportunistic, than Altman’s (a view I expressed, for example, in this podcast).
With that in mind, take a look at how ChatGPT allocated the silver medals for Chinaphobia fomentation and Regulatory parasitism: