Sam Altman, Jensen Huang, and Anthropic executives asked G20 ministers to reject binding AI rules, while Google shipped Gemini 3.8 Flash at sharp price cuts.
OpenAI CEO Sam Altman stood beside US Commerce Secretary Howard Lutnick in North Carolina on Wednesday and told trade ministers from the world's largest economies that adopting artificial intelligence is "non-negotiable." He compared the technology to electricity and urged governments to respect each other's sovereignty over how they deploy it.
The setting was the G20 Innovation Ministerial, hosted by Lutnick at a US venue that left little ambiguity about the message. Altman, Nvidia CEO Jensen Huang, and an Anthropic executive were all on hand to argue against binding AI rules. The lobbying push comes at a moment when the European Union is tightening its grip on US frontier labs (Gizmodo).
The coordinated appearance marked an unusually direct alignment between America's three most consequential AI companies and the Trump administration on global AI policy. It also came just two days after the EU formally classified ChatGPT as a "very large online search engine," a designation that pulls OpenAI into a stricter compliance regime under the bloc's Digital Services Act.
The market reaction
Huang took the earliest slot, sharing a fireside chat with Lutnick in which he asked officials to refrain from regulating "hypothetical, theoretical harm." Companies, he argued, should "take it upon themselves to develop the technology safely." Altman's pitch leaned economic. The benefits of AI, he said, will be "too high to ignore," and predicted the current wave would produce "the greatest boom in entrepreneurship and the creation and development of small businesses the world has ever seen."
That optimism sits uneasily alongside OpenAI's recent track record. The company has shipped its ad business early across seven markets while pulling back from owning checkout, releasing no GPT-6, and slipping its first consumer device to 2027, a pattern that suggests the parts of the roadmap that monetise attention arrived ahead of schedule and the parts requiring OpenAI to own transactions or hardware slipped (Index Lab). Altman himself once called mixing ads with AI "uniquely unsettling" at Harvard in May 2024.
Anthropic's footprint at the G20 was smaller but its commercial momentum is not. On August 26, Salesforce announced "Claudeforce," a strategic partnership that bakes Claude into the CRM giant's product suite. Salesforce plans to spend roughly $300 million on Anthropic tokens in 2026, on top of an existing equity stake valued near $5 billion. Claude becomes the default reasoning engine behind Salesforce's Agentforce suite, with 37 prebuilt sales skills shipping in a product called "Salesforce in Claude" (Crypto Briefing).
What the ministers heard
The CEOs' arguments landed in a regulatory environment that is fragmenting rather than converging. In the US, the Trump administration has signalled it prefers voluntary commitments and executive-branch guidance over legislation. In Brussels, frontier models trigger systemic-risk obligations under Article 52 of the EU AI Act, with a two-week window between training and notification, a window shorter than the gap between recent Google Flash releases.
Google, notably absent from the G20 lobbying stage, was shipping its third Flash model in six weeks. Gemini 3.8 Flash launched on Wednesday with introductory pricing of $0.75 per million input tokens and $3.75 per million output tokens, both set to roughly double after 2026 (The Verge). Artificial Analysis measured output tokens per task rising 30 percent and total cost up roughly 40 percent against the 3.7 release despite unchanged per-token pricing, because the new model "calls tools" more aggressively.
Google also released Gemini 3.8 Flash Cyber, a security-tuned variant that the Chrome Security team said fixed vulnerabilities 2.6 times more often than the best larger commercial models (News Bytes). Flash Cyber is restricted to trusted testers and governments, a distribution channel that places the model outside the EU's general-purpose AI documentation regime but raises its own questions about which governments get early access.
The implicit argument
The G20 lobbying and the Google price cuts describe the same strategy from different ends. US labs want the rules written by the country where they are headquartered, where voluntary commitments dominate. Hyperscalers want to win the price war so thoroughly that compliance costs elsewhere become a competitive nuisance rather than a moat.
That calculus leaves European officials with limited leverage. Article 53 requires every general-purpose model placed on the EU market to ship with technical documentation, a copyright policy, and a public training-data summary. Google signed up willingly on 30 July 2025, a week after Meta refused (The Next Web). The harder question is whether the EU's notification window can survive a release cadence measured in weeks.
The ministers in North Carolina will return home to budgets, elections, and AI strategies shaped far more by domestic politics than by one fireside chat. Whether the US industry's preferred model, light-touch, company-led, sovereignty-respecting, becomes the global default will depend less on what Altman and Huang said this week than on whether Brussels and Beijing decide to enforce their own clocks.
FAQ
What did Sam Altman say at the G20 Innovation Ministerial?
Altman said adopting the technology is "non-negotiable," compared it to electricity, and told ministers OpenAI wants to "respect all of your sovereignty" over how AI is used.
Why is OpenAI under regulatory pressure in Europe?
The EU reclassified ChatGPT as a "very large online search engine" on Monday, pulling it into stricter obligations under the Digital Services Act.
How much is Salesforce paying Anthropic?
Salesforce plans to spend roughly $300 million on Anthropic tokens in 2026, on top of an equity stake valued near $5 billion.
What is Gemini 3.8 Flash and why does the price matter?
It is Google's third Flash release in six weeks, priced at $0.75 per million input tokens and $3.75 per million output tokens until the end of 2026, after which both rates roughly double.
