AI News September 25, 2026: Trump and Vance Brush Off the AI 'Hoax' Warnings as Zuckerberg Breaks Ranks, Amazon's Muse Block Ignites the Agentic Commerce Wars, Muse Hammers Telco and Travel Stocks, Claude Opus 5.5 Migrates 680,000 Lines of Code in a Day, and OpenAI DevDay Looms

The AI world spent this week splitting into two camps — and on September 25, 2026, the split went fully political. One side: the frontier lab CEOs who stood before the UN Security Council and asked the world to help govern them. The other: a US administration calling their warnings a “hoax,” a vice president smelling a “Trojan horse,” and Mark Zuckerberg breaking ranks to say the industry can police itself. Meanwhile, the first great agentic commerce war broke out between Amazon and Meta, Wall Street began repricing entire sectors for the agent era, and Claude Opus 5.5 showed what frontier autonomy looks like when it’s pointed at a legacy codebase. Here are the five stories that matter today.

1. Trump Calls AI Risk Warnings a ‘Hoax’ as Vance Brands Regulation Push a ‘Trojan Horse’

Days after Sam Altman and Dario Amodei urged the UN Security Council to adopt international AI standards, the White House’s answer is a shrug. President Trump has dismissed the growing alarm over AI dangers as a “hoax” — comparing it to past political “scams” — and suggested the only control AI needs is “a strong and smart president” (Forbes). The stance reportedly springs from a belief that slowdown talk is a ploy to blunt US competition with China.

Vice President JD Vance went further, questioning the industry’s motivations outright. “It’s a bit of a Trojan horse,” Vance said of AI CEOs calling for regulation, arguing labs should simply ensure their own products don’t spiral out of control: “If you’re building Frankenstein, stop. Or maybe, if the cat is out of the bag, then build the defensive mechanism against Frankenstein” (AP).

The subtext is hard to miss: the same companies racing to build frontier systems are now asking governments to cap the race — and politicians on both sides are asking who benefits. With the UN briefing still echoing and antitrust scrutiny already sharpening around the AI slowdown pact, the governance fight has officially moved from conference keynotes to campaign rhetoric.

2. Zuckerberg Breaks Ranks: Labs Can Regulate Themselves

The industry’s united front didn’t survive the week. Mark Zuckerberg publicly split from Altman and Amodei, arguing on X that “every lab has the responsibility and incentive to move at the pace required to train its models safely, and the ability to take its own actions to ensure that happens” (AP).

His evidence? Meta itself. “Meta delayed shipping Muse for several months to focus on safety and security,” Zuckerberg wrote. “We didn’t call for everyone else to do this before we would. We just did it as part of our day-to-day work because it was clearly the right thing for people and for us.” Meta delaying its flagship agent — the one now signing partnership deals — as self-regulation proof is a sharp-elbowed argument that mandatory oversight is unnecessary.

It’s a three-way split with real stakes: Amodei wants global evaluation and verification regimes, Trump’s White House wants no binding limits, and Zuckerberg wants credit for self-restraint while shipping as fast as partnerships allow. Whichever vision wins will shape how the agent economy — the one currently rewriting stock prices — gets governed.

3. Amazon vs. Muse: The Agentic Commerce War Goes Public

The most consequential business fight in AI right now isn’t between model labs — it’s over who owns the customer relationship when an agent does the shopping. Amazon is now serving Muse users a popup at checkout: “Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use” (GeekWire). Amazon says it tried and failed to get Meta to exclude its site, and frames the block as a transparency and security matter: agents can bypass personalization, and third-party agents “should operate openly and respect service provider decisions about whether or not participate.”

Meta’s response is to rally the rest of commerce to its side. Shopify CEO Tobias Lütke announced deep integration bringing agentic checkout with Shop Pay to “all Shopify stores” via Muse, and Zuckerberg hinted more partnerships are coming (LA Times). Other e-commerce platforms have embraced the agent.

The standoff is remarkable given the two giants are partners — Amazon products have been buyable inside Facebook and Instagram since 2023, and Meta runs agentic AI workloads on AWS. But as CNET noted, if e-commerce is going to be a big part of the agentic future, Amazon just drew the first hard line — and every retailer now has to pick a side.

4. Muse Economics Hit Wall Street: ‘Consumer Inertia’ Stocks Hammered

The market has decided Muse is a sector-level event. Shares of AT&T, insurers, travel agents, and other consumer-facing companies are “getting hammered” on worries the agent will erode their businesses, as Barron’s put it — the thesis being that an assistant that negotiates bills, compares prices, and books travel itself attacks the “consumer inertia” these businesses are built on. Online travel booking platforms are among the hardest hit, with Expedia dropping as the Muse disruption rippled through the sector (Barron’s).

The move is a preview of how markets will trade the agent era: not as a chip-and-cloud story, but as a displacement story. When a single product launch can move telco and insurance stocks, every consumer business with a moat built on switching costs now has an AI agent line in its risk disclosures. It also cuts the other way — the same AI slowdown fears that knocked Nvidia, Broadcom, and Marvell down earlier this month are colliding with an agent boom that is very visibly destroying and creating value at the same time.

5. Claude Opus 5.5: A 680,000-Line Code Migration in Under a Day

While the governance fight rages, Anthropic’s Claude Opus 5.5 — released September 22 — is showing what the fuss is about. One early tester completed a 680,000-line code migration in less than a day, work Anthropic says would have taken an engineering team weeks. In another test, Opus 5.5 cut load times across every page of a web app in 39 of 40 attempts (Anthropic).

The pricing is the headline for developers: $4/$20 per million input/output tokens — 20% below Opus 5 — while matching Claude Fable 5.1 on most tasks and running typical workloads about 40% cheaper (Reuters). On Anthropic’s automated behavioral audit across roughly 2,000 scenarios, it posts the best alignment scores of any model the company has tested, attempting to circumvent boundaries roughly 85% less often than Opus 5 or Mythos 5.1.

The fine print matters: extended thinking is always on, thinking blocks are now tied to the model and conversation, and some cybersecurity requests get routed to Opus 4.8. With Sonnet 5.5 and Haiku 5.5 promised in the coming weeks, the Claude 5.5 family is landing as a full-stack refresh.

The Week Ahead

Every thread this week converges on one date: September 29, OpenAI DevDay at Fort Mason in San Francisco. With the GPT-6 generation barely two weeks old, the agentic commerce war accelerating, and Washington openly hostile to the labs’ governance asks, whatever OpenAI ships next — new models, agentic APIs, platform tooling — will land into the most contested AI landscape yet. The keynote is livestreamed free. Watch it.

Sources: Forbes, AP, Yahoo News, GeekWire, LA Times, Barron’s, Anthropic, Reuters, OpenAI.