🧯 Trump cools AI rules | Axios AI+ Summary

1. Key Themes

Theme 1: U.S. AI Regulation Is Deliberately Staying Weak to Protect Competitiveness

The Trump administration is choosing a voluntary, industry-friendly framework over mandatory oversight — explicitly to avoid hampering U.S. AI companies.

"The surprise move comes more than a week after Trump canceled the release of another version of the order with stricter requirements, saying it could have hurt American competitiveness."

"Sacks was able to secure a shorter window for pre-deployment testing — 30 days — and a voluntary framework as some pushed to make it mandatory."

Theme 2: Token Consumption Is Exploding — and Cost Has Become the #1 Enterprise Pain Point

Usage growth is so extreme it has created a brand-new category of problem for enterprise buyers, almost overnight.

"The token leader at OpenAI uses about 100 billion tokens a month... Over six years ago, the top token user at OpenAI blew through about 100,000 tokens per month." — Sam Altman

"Cost concerns are the second-most common issue he hears about from customers behind simplifying AI workflows. Cost never came up before but is now 'all of a sudden a huge issue.'"

Theme 3: AI Agents Are Moving Into Commerce and Productivity Workflows

AI is graduating from answering questions to taking autonomous actions — filling shopping carts, preparing meeting briefs, and running proactively in the background.

"Gopuff is launching 'Go,' an AI shopping assistant that works via voice and text and can automatically add items to your cart based on its memory and context about your previous purchases."

"What Altman said is coming next from OpenAI: 'constant running proactive AI.'"

Theme 4: Big Tech Is Racing to Own the Full AI Stack — Independent of Any Single Partner

Microsoft's Build announcements signal a deliberate push to reduce dependence on OpenAI and control its own models, data, OS, and hardware.

"Microsoft is seeking to show it is a serious player in AI beyond its relationship with OpenAI."

"Microsoft also announced Project Solara, an Android-based operating system designed to run AI agents on a range of small devices such as earbuds and speakers."

Theme 5: AI Cybersecurity Threats Are Severe Enough to Force Government Action

The executive order was triggered in part by AI-exposed security vulnerabilities, signaling a new national security dimension to AI risk.

"The new order lets the White House kick the can down the road while it considers new rules for cutting-edge AI models and what to do about AI's advanced cybersecurity capabilities."

"The order is an attempt to shore up the country's cyber defenses as models like Mythos reveal shocking cybersecurity vulnerabilities."

2. Contrarian Perspectives

Perspective 1: The "Light-Touch" AI Regulation Framing May Be a Competitive Risk Dressed as an Opportunity

The administration framed loosening oversight as pro-competitiveness, but the same order acknowledges that advanced AI introduces serious national security vulnerabilities — meaning deregulation may be creating the very risks it claims to avoid.

"Advanced AI capabilities make our Nation stronger, but also introduce new national security considerations that require coordinated action."

"The order is an attempt to shore up the country's cyber defenses as models like Mythos reveal shocking cybersecurity vulnerabilities, but it doesn't compel AI companies to share information about their latest models."

The government is simultaneously acknowledging dangerous new attack surfaces while removing the mandatory disclosure mechanisms that would allow it to assess them.

Perspective 2: AI Cost Will Constrain Adoption Before Capability Does

The conventional narrative is that AI capability is the binding constraint on enterprise adoption. But the evidence here suggests cost has already emerged as the #1 friction — even before "constant running proactive AI" hits the market.

"Cost never came up before but is now 'all of a sudden a huge issue.'"

"Uber put a $1,500 per month limit on AI coding tools after multiple executives had viral comments about token usage."

If leading enterprises are already capping usage, the path to AI ubiquity runs through cost compression, not capability advancement.

Perspective 3: Microsoft Betting on Commercially Licensed Training Data Could Become a Major Differentiator

While most AI models rely on broadly scraped "publicly available" data, Microsoft has specifically positioned MAI-Thinking-1 as trained only on licensed data — a potential legal and enterprise-trust moat as IP litigation heats up.

"Microsoft also noted that it was not distilled from any other models and is trained only on commercially licensed data, rather than simply the kinds of 'publicly available' data typically used to train large language models."

This positioning could prove strategically valuable for risk-averse enterprise and government buyers facing growing scrutiny over training data provenance.

3. Companies Identified

OpenAI Description: Leading AI lab and maker of ChatGPT Why mentioned: Top token usage data reveals explosive enterprise consumption growth; cost has emerged as a top enterprise concern; CEO signals "constant running proactive AI" as the next product direction

"The token leader at OpenAI uses about 100 billion tokens a month... That's a 1 million-fold increase in token usage." "ChatGPT hit 1 billion monthly active users in just three years, setting a new record for the fastest app to reach the milestone."

Microsoft Description: Enterprise software and cloud giant Why mentioned: Debuted its first internally developed reasoning model (MAI-Thinking-1), a personal AI agent (Scout), and Project Solara — all positioning the company as an independent AI player beyond its OpenAI partnership

"Microsoft is seeking to show it is a serious player in AI beyond its relationship with OpenAI." "Microsoft also announced Project Solara, an Android-based operating system designed to run AI agents on a range of small devices such as earbuds and speakers."

Gopuff Description: Instant delivery platform Why mentioned: First case study of a major retailer deploying an agentic shopping assistant (Go) powered by Grok, pointing to AI-commerce as a new revenue model

"Gopuff is launching 'Go,' an AI shopping assistant that works via voice and text and can automatically add items to your cart based on its memory and context about your previous purchases."

SpaceXAI (xAI) Description: Elon Musk's AI company, maker of Grok Why mentioned: Grok is the AI backbone powering Gopuff's new agentic commerce product, establishing xAI as an emerging enterprise API competitor

"Gopuff... is launching an agentic personal shopping assistant that runs on Elon Musk's SpaceXAI chatbot Grok, turning it into a tool that fills your cart for you."

Uber Description: Ride-sharing and delivery giant Why mentioned: Real-world example of enterprise AI cost controls — imposed a hard $1,500/month cap on AI coding tools due to runaway token spend

"Uber put a $1,500 per month limit on AI coding tools like Claude Code to cut costs." (Bloomberg, cited in article)

Anthropic Description: AI safety-focused lab, maker of Claude Why mentioned: IPO storylines flagged as a key near-term market event to watch

"Five Anthropic IPO storylines to watch." (Axios, cited in article)

Meta Description: Social media and AI conglomerate Why mentioned: Scaled back a controversial plan to track employee mouse movements to generate AI training data, following internal pushback

"Meta is reining in its plans to track employee mouse movements to train its AI models, due to concerns from staff."

4. People Identified

Sam Altman Description: CEO of OpenAI Why mentioned: Disclosed the staggering scale of OpenAI's top enterprise token user and publicly named cost as a top customer concern; previewed "constant running proactive AI" as OpenAI's next frontier

"The token leader at OpenAI uses about 100 billion tokens a month. To my embarrassment, that's not the token leader in the world. We found someone that used even more." "We want you all to be able to use AI and never worry about it being great and affordable."

David Sacks Description: Former White House AI czar, current adviser Why mentioned: Key behind-the-scenes architect of the lighter-touch AI executive order — successfully blocked mandatory licensing and shortened pre-deployment testing windows

"Sacks was able to secure a shorter window for pre-deployment testing — 30 days — and a voluntary framework as some pushed to make it mandatory."

Ryan Baasch Description: National Economic Council deputy director Why mentioned: Worked alongside Sacks to insert anti-mandatory licensing language into the executive order

"Former White House AI czar and current adviser David Sacks and National Economic Council deputy director Ryan Baasch pushed for language prohibiting the creation of mandatory government licensing."

5. Operating Insights

Insight 1: Enterprises Must Build AI Cost Governance Now, Before Usage Scales

Uber's $1,500/month cap — imposed reactively after executives had "viral comments about token usage" — illustrates that enterprises without proactive cost guardrails will face runaway spend as agentic AI usage compounds. Operators should instrument token usage by team and use case, set tiered budgets, and model cost trajectories before deploying autonomous agents at scale.

"Cost never came up before but is now 'all of a sudden a huge issue.'"

Insight 2: The Winning AI Product Design Pattern Is Contextual Memory + Situational Awareness

Gopuff's "Go" agent doesn't just respond to queries — it proactively predicts needs, monitors consumption patterns, and reads environmental signals like weather and local trends. This points to a product design principle: agents that hold user context over time and act on it unprompted will have materially higher retention and monetization potential than reactive chatbots.

"Rather than searching for specific items, users can describe a situation, like a game-day party they're hosting or the desire for a healthy breakfast, and the AI agent will assemble a cart automatically based on that prompt." "Go can also predict when shoppers are running low on items like coffee or paper towels, which it says it can pack and deliver from its own warehouses in as little as 15 minutes."

6. Overlooked Insights

Insight 1: A "Cybersecurity Clearinghouse" Could Become a Quiet Regulatory Chokepoint

Buried in the executive order is a mandate for a "cybersecurity clearinghouse" and a classified benchmarking process to assess AI models' cyber capabilities — with a 60-day deadline involving NSA, CISA, NIST, and Treasury. Though framed as voluntary today, this infrastructure creates the institutional foundation for mandatory reporting in a future administration. Investors in frontier AI companies should monitor whether this quietly becomes a de facto disclosure regime.

"Within 60 days, the Treasury Department, the National Security Agency, the Cybersecurity and Infrastructure Security Agency, the National Institute of Standards and Technology and White House officials must 'develop and maintain a classified benchmarking process to assess the advanced cyber capabilities of AI models' and decide when a model should be treated as a 'covered frontier model.'"

Insight 2: Microsoft's "Commercially Licensed Data Only" Claim on MAI-Thinking-1 Is a Potential IP Liability Shield

Mentioned briefly as a technical footnote, the fact that MAI-Thinking-1 is trained exclusively on commercially licensed data — not scraped web content — could become a significant enterprise procurement differentiator as AI copyright litigation accelerates globally. This is a positioning move as much as a technical one.

"Microsoft also noted that it was not distilled from any other models and is trained only on commercially licensed data, rather than simply the kinds of 'publicly available' data typically used to train large language models."