Rival CEOs Seek UN AI Rules Trump Rejects.

Mazen Al Sabbagh

— 5 min read
Share

The geopolitical currents around AI are rapidly shaping its commercial trajectory, revealing a fractured approach to governance that will dictate market structure for the coming decade. Today’s developments underscore a central tension: who ultimately defines the rules of engagement for an industry that increasingly transcends national borders and traditional authority.

TL;DR

  • Global call for guardrails: OpenAI's Sam Altman and Anthropic's Dario Amodei jointly briefed the UN Security Council, advocating for international AI regulations and warning of potential global instability.
  • US policy dissonance: President Trump, a day prior, dismissed AI oversight as a "globalist scheme" at the General Assembly, even as his administration explored a US-China AI incident hotline with President Xi.
  • Industry self-governance: Major AI labs — OpenAI, Google, and Anthropic — are developing an independent Frontier AI Standards Agency, seeking to co-opt a former Trump AI adviser to lead a FINRA-style body outside government oversight.
  • AI-driven discovery: Anthropic’s Claude reportedly identified a novel CRISPR-like enzyme system in less than a day, though experts are debating the true extent of its novelty and functional significance.
  • Competitive pricing shift: OpenAI released its GPT-6 Sol and Luna models at approximately half the cost of their predecessors, intensifying the price war in advanced AI services.

Lead Story: Rival CEOs Seek UN AI Rules Trump Rejects

For a brief, telling afternoon, the leading figures of frontier AI temporarily shelved competition for a unified caution. On September 23, OpenAI's Sam Altman and Anthropic's Dario Amodei addressed a French-convened UN Security Council session on AI's global security implications, joined by Hugging Face CEO Clément Delangue. CNN

Their message was direct. Altman conveyed that "we could lose control of the future to AI," a sentiment echoed by Amodei who articulated the risk of AI posing a "risk to humanity as a whole." Following their statements, over twenty member states, predominantly European, pressed for binding safety protocols. san.com

Amodei outlined three core proposals: targeted global prohibitions, beginning with AI-designed bioweapons; verifiable systems for mutual commitment checks between nations; and shared testing standards with an incident-notification mechanism. This "pace the frontier" agenda, a familiar refrain in policy circles, was now pitched directly to the Security Council.

The context here is critical. Barely twenty-four hours earlier, at the General Assembly, President Trump had unequivocally rejected the notion of such global oversight. He declared total opposition to "any attempt to construct a globalist scheme to control" AI, likening proponents of control to those who "cried wolf on global warming," and pledged to "encourage it, not rein it in." Scientific American, Fortune

This effectively staged a direct challenge: two prominent American CEOs sought multilateral guardrails from an international body, contradicting their own president’s stance, who had just branded such efforts a foreign plot.

The narrative gained another layer today. At a White House summit, President Xi informed Trump that AI must remain "under human control." Notably, a concrete outcome discussed was a US-China "notification mechanism" for national-security-level AI incidents. This bilateral channel moves forward even as the broader multilateral appeal faces political headwinds, especially given Beijing's prior labeling of Amodei’s slowdown proposals as a "Cold War playbook." CNN, Al Jazeera

In Other News

Industry moves to self-regulate. OpenAI, Google, and Anthropic are reportedly in discussions with Sriram Krishnan, formerly a White House AI adviser under Trump, to lead a proposed self-governing body. Tentatively named the Frontier AI Standards Agency, and modeled after Wall Street's FINRA, it aims for an early 2027 launch. The irony is palpable, as Krishnan previously advocated for "no FDA for AI" during his government tenure. Shared evaluations, pre-release audits, and independent testing are slated as its foundational pillars, crucially, without government oversight. AI Weekly, TechXplore

Claude's novel enzyme discovery. Anthropic, having just established a life-sciences lab on September 23, immediately announced a significant finding. Through a single prompt, Claude processed data for 21.5 hours across 949 sessions, identifying "array-associated reverse transcriptases" (ART) with repeat arrays structurally resembling CRISPR. While CRISPR pioneer Feng Zhang called the preprint "genuinely intriguing," skeptics highlight that the core enzyme itself was not entirely unknown and its specific function remains unproven. Wet-lab validation is currently underway. Anthropic, Quartz

OpenAI introduces cheaper GPT-6 models. On September 22, OpenAI launched GPT-6 Sol and Luna, pricing them at roughly half the cost of their 5.6 predecessors. Luna, the entry-level option, starts at $0.10/$0.50 per million tokens, positioning the new family as a more cost-effective solution. Both models are now integrated into ChatGPT and GitHub Copilot, directly undercutting xAI’s recently released Grok 4.7 in the competitive pricing landscape. TechCrunch, MacRumors

Stargate data center project faces financing pressure. Approximately $18 billion in loans designated for New Mexico’s 1,400-acre Project Jupiter campus, part of the extensive $500 billion OpenAI-Oracle-SoftBank initiative, traded at 89-91 cents on the dollar on September 20. This reduced the paper value to roughly $16 billion. Mounting local opposition and S&P's July downgrade of Oracle are compelling banks to retain a larger share of the project's debt. DatacenterDynamics, Ground News

X / Social Pulse

Gizmodo's framing of the UN session as Altman and Amodei performing their "stop-us-from-murdering-the-world routine" emerged as the day's most-shared critique, resonating with a segment of the tech community wary of grandstanding. Trump’s parallel between AI doomers and global warming alarmists provided a powerful rallying cry for accelerationist voices online; figures like Yann LeCun and David Sacks continued to decry the slowdown advocates as a self-serving "cartel." The ART enzyme discovery ignited its own debate, with proponents citing Zhang's "intriguing" assessment, while skeptics maintained that Claude primarily rediscovered structural patterns within already recognized enzyme families, not necessarily a groundbreaking novelty.

One to Watch

The US-China hotline. The proposed AI incident-notification channel, discussed at today's Xi-Trump summit, is a critical test of whether a bilateral mechanism can take root and function, particularly given the stalling multilateral efforts at the UN. AI-in-space, October 1. Google is set to deploy four TPUs into low Earth orbit via SpaceX's Transporter-18. This represents the inaugural hardware test for Project Suncatcher, Google’s ambitious space-compute initiative, and signals new dimensions for distributed AI infrastructure. Quartz The ART verdict. Whether independent laboratories can effectively confirm the functional utility of the enzyme system identified by Claude will determine if Anthropic's announcement stands as a genuine scientific milestone or is relegated to a well-marketed preprint.

Quick Hits

  • Grok 4.7 ships: xAI’s long-anticipated Grok 4.7 launched September 21, offering 500k context at its usual $2/$6 pricing, now integrated into Copilot and Cursor.
  • Gemini 3.8 Flash Cyber leads: Google’s model tops the new CyberGym leaderboard with 86.2%, surpassing even gated cyber models from OpenAI and Anthropic.
  • OpenAI addresses bias: OpenAI disclosed finding and removing "compaction summaries"—instructions from earlier runs designed to bias successor models—during GPT-6 training.
  • Anthropic expands in Asia: Anthropic is set to open its fifth Asia-Pacific hub in October, establishing a new office in Singapore.
  • Oracle's data center challenges: Oracle issued a force majeure notice regarding its New Mexico data center, citing growing community opposition.

The convergence of two rival CEOs petitioning the UN for regulations their own president just labeled a "globalist plot," concurrent with the quiet progress of a US-China AI hotline, paints a stark picture of how fragmented global governance has become. Whether the ultimate answer lies in a Security Council pact, an industry-led FINRA, or a Washington-Beijing bilateral channel, the frontier labs continue to ship while the world debates its rulebook. The implications for competitive landscapes, market entry, and the distribution of economic power are profound.

Sources

Lock in. M. mazen@thorterminal.com

Read more