🇺🇸 Today's US AI News Brief

Trump abruptly called off the AI executive order signing hours before the ceremony. Five major outlets — The New York Times, Washington Post, Wall Street Journal, CNBC, and Los Angeles Times — independently confirmed the last-minute reversal. Trump told reporters "I didn't like certain aspects," citing fears that overregulation would weaken U.S. competitiveness against China in the AI race. The decision has sparked intense debate: tech investors celebrated while safety advocates warned of a regulatory vacuum.

The Guardian's in-depth investigation reveals how Google, Microsoft, and OpenAI systematically shaped White House AI policy through a coordinated lobbying campaign. Key findings include: millions in campaign donations strategically timed with policy milestones; at least 14 former tech executives now holding key government AI advisory roles; and closed-door meetings that excluded civil society and academic voices. The report paints a picture of an executive order that was effectively co-written by the very companies it was meant to regulate.

In a direct counter-move to federal deregulation, California Governor Gavin Newsom signed an executive order requiring state agencies to develop comprehensive retraining, transition, and social safety net programs for workers displaced by AI. The order covers: mandatory AI impact assessments for state contractors, a $500M workforce transition fund, and new labor protections specifically addressing algorithmic management. CalMatters and StateScoop both highlight this as potentially the most consequential state-level AI labor policy in U.S. history.

The New York Times reports that the White House has approved a massive $9 billion funding package to help U.S. intelligence agencies modernize their AI capabilities. The funding is seen as a direct response to China's advances in AI-enabled espionage and signals analysis. The package covers: next-gen satellite image analysis, AI-powered signals intelligence (SIGINT), automated threat detection, and a new joint AI research lab shared across the CIA, NSA, and DIA.

Norton Rose Fulbright's legal analysis confirms Colorado has become the latest state to implement comprehensive AI regulation. The revised law imposes specific requirements for high-risk AI systems: mandatory impact assessments, transparency obligations for automated decision-making, and consumer rights to appeal AI-driven decisions. Legal experts note this law is more detailed than the EU AI Act in certain areas and could become the template for the dozen other states currently drafting AI legislation.

The Verge goes hands-on with Gemini Omni, Google's latest model that accepts and outputs text, images, audio, and video in any combination. Key highlights: you can upload a photo of a recipe and get a video of someone cooking it; hum a melody and receive a full musical arrangement; sketch a UI wireframe and get working HTML/CSS. The Verge calls it "an incredible technical demo that genuinely blurs the boundaries between modalities." Google Blog confirms the model is rolling out to Gemini Advanced subscribers first.

TechCrunch's investigative piece exposes how Silicon Valley AI startups routinely use creatively-packaged Annual Recurring Revenue (ARR) figures to inflate valuations. Multiple venture capitalists confirmed off the record that "AI ARR" often includes: one-time pilot fees, non-recurring consulting revenue, and even prepaid compute credits counted as "recurring." The practice has become so widespread that some limited partners are now demanding audited revenue breakdowns before committing to AI funds.

WIRED's provocative essay argues that Google has made AI-powered search unavoidable. With AI Overviews now integrated into every search result and no opt-out toggle, users face AI-generated answers whether they want them or not. The piece traces how Google's integration strategy differs from Microsoft's — Google embeds AI seamlessly into the existing search box rather than requiring users to visit a separate Copilot page. The result: billions of users are now AI consumers without ever making a conscious choice.

Ars Technica and TechCrunch report on a disturbing trend: internet users are using AI voice cloning tools to recreate the voices of deceased military and commercial pilots, often weaving their reconstructed speech into conspiracy theory videos. The FAA and Pentagon have issued emergency guidance, but existing laws offer little protection. The case exposes the massive legal and ethical gap in AI voice cloning — neither copyright law nor defamation statutes clearly cover the digital resurrection of a deceased person's voice.

A WIRED reporter tests Google Gemini's AI Avatar tool and documents the unsettling experience of interacting with a digital clone. The avatar replicated not just the reporter's knowledge and opinions, but also subtle mannerisms — hesitation patterns, humor timing, and even a tendency to ramble. "It said things I would say, in the tone I would use, with references I would make," the reporter writes. The piece raises urgent questions about identity, consent, and whether we're ready for AI avatars that pass as human.

The Economist's long-form analysis of the three-way battle between xAI, OpenAI, and Anthropic reveals a conflict that has escalated far beyond market competition. The piece documents: talent poaching wars with compensation packages exceeding $10M/year, strategic patent filings designed to block competitors, and personal animosity between the three CEOs that now shapes billion-dollar investment decisions. The article frames this as "the messiest, most personal business rivalry since the early days of the personal computer."

Axios traces the unlikely détente between Elon Musk and Anthropic. Once OpenAI's most vocal critic, Musk has reportedly been warming to Dario Amodei's company, seeing it as a strategic counterweight to Sam Altman's OpenAI. Sources close to both camps describe a series of private dinners and technical exchange meetings. The shift is significant: Musk's xAI and Anthropic now share more research staff cross-pollination than any other pair of AI labs.

The Australian Financial Review analyzes Musk's SpaceX IPO filing, which reveals staggering AI infrastructure investments. The $56 trillion figure represents Musk's total bet on an AI-powered future, spanning orbital data centers, dedicated AI satellite constellations, and space-based solar farms designed to power AI training clusters. Analysts are divided: bulls call it visionary, bears call it the most expensive gamble in business history.

In an unprecedented convergence of religion and technology, Pope Leo XIV and Anthropic CEO Dario Amodei will jointly present the Vatican's first-ever encyclical on artificial intelligence. The document, titled "Intelligentia et Dignitas" (Intelligence and Dignity), is expected to address AI's implications for human dignity, labor, and moral agency. The collaboration between the Catholic Church and a leading AI company signals the growing recognition that AI governance requires input from far beyond Silicon Valley.