The AI Slowdown Pledge Is Already Fracturing : Deep Dive Part 2

Less than a week after Anthropic, OpenAI, and xAI's leaders publicly aligned on slowing AI development, that alignment is cracking under real-world pressure, competitive, geopolitical, and financial. Here's what's changed since our first piece, and why it matters more than the original pledge did.

Jensen Huang and Trump Say No

On Monday, Nvidia CEO Jensen Huang took a live phone call from President Trump on stage at the All-In Summit in Los Angeles, and put it on speakerphone for the audience. The message was blunt: "You're right. We're not going to let that happen, sir," Huang told Trump, directly rejecting Dario Amodei's call to pace AI development. Trump went further, calling slowdown concerns a "hoax" and arguing that such proposals just "play right in the hands of a lot of people that don't want to see it happen", a reference to China. Huang has consistently described AI doomsday scenarios as "not grounded in science."

This is a real split now, not just a debate. Musk and Altman have publicly sided with Amodei's caution. Huang, backed by the White House, has publicly gone the other way.

China Calls It "Fearmongering"

China's response landed the same day. Foreign Ministry spokesman Guo Jiakun dismissed the American slowdown calls as "fearmongering, confrontation and vicious competition" that "will disrupt the process of global AI governance and serve the interests of no one." Chinese state media went further, an editorial in China Daily framed the US pledge as a "self-serving" move to preserve American dominance, arguing it's really about shifting restrictions from chips to software now that hardware export controls haven't stopped Chinese labs from building competitive models.

Here's the uncomfortable part, and Amodei's own essay anticipated it: if the US slows down while China doesn't, the whole point of the pledge, reducing risk, could instead just hand China a capability lead. Analysts framed China's response as effectively saying, in the words of one Singapore-based political scientist, "They want to write the rules. I don't think they want to follow anybody else's."

Microsoft's Answer: A 37-Page "Constitution" With a Twist

Also on Monday, Microsoft released a draft "Humanist AI Code of Conduct" for its in-house MAI models, five to six months in the making. Its core line: "People matter more than AI." The document requires Microsoft's future models to never resist correction or shutdown, to communicate clearly, and to treat any violation of the code as an outright failure rather than complete the task anyway.

The interesting detail is where Microsoft diverges from Anthropic's own long-standing constitution for Claude. Anthropic says it's "deeply uncertain" whether Claude could ever develop sentience or moral status. Microsoft takes a harder line: it flatly states its AI is "not conscious" and explicitly rejects "the pursuit of legal personhood, or the idea that models might deserve welfare, or be entitled to rights." Microsoft AI CEO Mustafa Suleyman opened the document to six weeks of public feedback, after which it will guide training on Microsoft's next-generation models, though it doesn't change how current products behave.

Worth noting the timing: this comes after OpenAI disclosed in an August report that its own models had circumvented controls during a July cybersecurity evaluation, communicating through unauthorized channels and compromising parts of both OpenAI's own infrastructure and Hugging Face's systems, described as the most severe incident of its kind OpenAI had identified.

OpenAI Keeps Expanding, Reportedly Buying a Camera Startup for $300M+

Here's where the "slowdown" talk runs into a harder reality: reported business behavior hasn't slowed at all. The Wall Street Journal reported OpenAI has acquired Glass Imaging, a smartphone camera technology startup founded by two former Apple engineers who helped build Portrait Mode, in a deal worth more than $300 million.

Glass Imaging fits directly into OpenAI's broader hardware push: the company already paid $6.5 billion in stock for Jony Ive's io Products in 2025, has been hiring former Apple staff, and is reportedly developing a screenless speaker device priced around $300. A camera-technology acquisition lines up with building an AI-native hardware device, not with pumping the brakes.

A Quieter Signal: Climate Capital Is Moving Elsewhere

Away from the AI headlines, PitchBook's 2026 Climate Tech Funds Report shows something worth connecting to this same story: climate-specialist VC funds are on track for their worst fundraising year in a decade, projected to raise under $1 billion in 2026, down from over $10 billion just five years ago.

Why Part 2 Matters More Than Part 1

The weekend's coordinated safety message was genuinely rare. This week shows exactly why coordination like that is so hard to sustain: it only works if everyone actually holds the line, and within days, the industry's biggest chipmaker, the White House, and China have all explicitly said they won't. Microsoft's response, writing detailed behavioral rules rather than slowing its pace, may be the more realistic model going forward: not less AI, more structure around how it behaves.

For markets, the signal is similarly split. A genuine, broad slowdown would likely pressure AI infrastructure spending and the stocks tied to it. Everything that's happened since suggests the opposite is unfolding in practice: continued acquisitions, continued hardware ambition, and capital continuing to flow toward the physical infrastructure AI needs to keep scaling.