🌀 AI rules scramble | AI+ Government Summary

1. Key Themes

Theme 1: AI Model Releases Now Require De Facto Government Approval

The Trump administration has established an informal but powerful vetting process for frontier AI models. Both OpenAI and Anthropic navigated government negotiations before their latest releases — a dynamic that "would have been unthinkable just months ago." The process involves "export control threats, licensing requirements and negotiations with a host of government agencies that are sometimes at odds." With a cybersecurity executive order now being implemented, "other AI labs are poised to face the same process."

Theme 2: The Absence of Regulatory Infrastructure Is Creating Avoidable Crises

The U.S. dismantled existing AI oversight mechanisms without replacing them, leaving no standardized framework when real safety issues emerged. The Biden-era AI executive order required companies to share safety testing results, including jailbreaking vulnerabilities — but "President Trump, vowing to pursue a deregulatory agenda on AI, scrapped that order's reporting requirements." When safety concerns with Anthropic's Fable model came to a head, "there was no alignment with industry and government on how severe jailbreaks need to be to raise a red flag." Experts argue that "had there been a framework to assess and standardize the severity of jailbreaking or safety bypassing, export controls may have been avoided."

Theme 3: The Government Lacks the Technical Capacity to Regulate AI Effectively

Structural underfunding and talent deficits mean the agencies nominally responsible for AI oversight cannot do their jobs. "Less than 1 percent of AI Ph.D.s go into government." The Center for AI Standards and Innovation (CAISI) — a key implementation body for Trump's AI action plan — has an "operational budget of $15 million, but it needs $84 million annually to fulfill Trump's AI action plan." Agencies like CISA and CAISI have been "sidelined" and underfunded, according to the Cato Institute's Kevin Frazier.

Theme 4: Global AI Governance Is Fragmenting Along National Lines

While the U.S. scrambles with ad hoc regulation, other jurisdictions are moving decisively. The UN opened its "first Global Dialogue on AI Governance in Geneva," with Secretary General Guterres declaring "Innovation needs guardrails … If AI is to be powerful, it must be governed." Simultaneously, China's rules on AI companions take effect July 15, with specific provisions banning virtual intimate relationships with minors and mandating age verification — prompting ByteDance and Alibaba to pull companion AI features ahead of the deadline. The rest of the world "has implemented tech privacy regulations, updated antitrust laws and passed transparency and research access measures," while "we didn't do any of it," per former Biden tech official Asad Ramzanali.

2. Contrarian Perspectives

Perspective 1: The Trump Administration's "Deregulatory" AI Stance Is a Myth in Practice

The conventional narrative is that Trump's AI policy is hands-off and pro-industry. In reality, AI companies face significant government pressure — just through informal and unpredictable channels rather than codified law. "Trump has not invoked the DPA, and the administration says the provisions of his cyber executive order are voluntary." Yet in practice, "it's clear that in today's regulatory environment, with what happened to Anthropic's Fable fresh in everyone's minds, AI companies need to keep the government happy." The deregulatory label masks a system that may actually be more unpredictable and riskier for companies than formal regulation would be.

Perspective 2: Biden's AI Order Was the Right Framework — Scrapping It Created the Problem It Was Supposed to Prevent

The Biden-era reporting requirement on jailbreaking was widely criticized as government overreach, but its absence is precisely what led to Anthropic's export control crisis. "The 'jailbreaking' issue was the type of vulnerability Amazon flagged last month that eventually led to export controls on Anthropic." A source familiar with the situation told Axios directly: "Had there been a framework to assess and standardize the severity of jailbreaking or safety bypassing, export controls may have been avoided." The very mechanism industry opposed would have shielded it from a far harsher outcome.

Perspective 3: State-Level AI Governance May Be the Most Practical Path Forward

With Congress failing to pass comprehensive AI safety legislation and the federal government underfunded and understaffed, state models may offer the most actionable blueprints. Kevin Frazier argues the government "could have learned from state initiatives, pointing to examples like Oklahoma's free AI literacy training, Utah's regulatory sandboxes for testing AI under heightened oversight, and Massachusetts' data practices." This bottom-up approach is non-obvious given how federally dominated the AI policy conversation has been.

3. Companies Identified

OpenAI

Description: Leading U.S. AI lab, maker of GPT models
Why mentioned: Successfully navigated government negotiations to secure a wide release of GPT 5.6; first major lab to go through the Trump administration's new informal vetting process
Quote: CEO Sam Altman: "I thought it was a very productive process. This was our first time through it, so there are things we'll learn about how to make it better next time, which we'll get going on soon." And: "You really want to be confident in your safety claims because otherwise the world is going to get uncomfortable very fast."

Anthropic

Description: AI safety-focused lab, maker of Claude models
Why mentioned: Its Fable model became the cautionary tale for the industry — subject to export controls after Amazon flagged a jailbreaking vulnerability, illustrating the stakes of the new regulatory environment
Quote: "The 'jailbreaking' issue was the type of vulnerability Amazon flagged last month that eventually led to export controls on Anthropic."

ByteDance

Description: Chinese tech conglomerate, parent of TikTok
Why mentioned: Pulling AI companion features ahead of China's July 15 regulatory deadline
Quote: Bloomberg reported that ByteDance and Alibaba are "pulling the plug on features that let users build and chat with AI companions" to prepare for these regulations.

Alibaba

Description: Chinese e-commerce and tech giant
Why mentioned: Also pulling AI companion features ahead of China's new rules
Quote: Same Bloomberg report noted Alibaba joining ByteDance in "pulling the plug on features that let users build and chat with AI companions."

4. People Identified

Sam Altman

Description: CEO of OpenAI
Why mentioned: Publicly characterized the government vetting process as productive while also signaling that safety credibility is now a business-critical requirement
Quote: "You really want to be confident in your safety claims because otherwise the world is going to get uncomfortable very fast."

Asad Ramzanali

Description: Former Biden administration tech official
Why mentioned: Argues the U.S. squandered years of opportunity to build regulatory foundations before AI became critical infrastructure
Quote: "We didn't do any of it… Given where things are, this is the right thing for the companies and for the government. But we should never have been here."

Kevin Frazier

Description: Cato Institute scholar and Director of the University of Texas' AI and Law Program
Why mentioned: Provides structural critique of government AI capacity and offers state-level governance as an alternative model
Quote: "The truth is that there's never been a 'right' answer for how to govern AI, but there are plenty of wrong ones," including deficient government expertise and failing to build trust among the public.

Rep. Josh Gottheimer (D-N.J.)

Description: U.S. Congressman
Why mentioned: Congressional voice criticizing the opacity and disorder of the current White House AI vetting process
Quote: "Right now, there is far too much confusion with the White House's AI vetting process — both for the country and for our leading AI developers."

António Guterres

Description: UN Secretary General
Why mentioned: Opened the first Global Dialogue on AI Governance in Geneva, calling for global rules prioritizing children's safety
Quote: "Innovation needs guardrails … If AI is to be powerful, it must be governed."

5. Operating Insights

Insight 1: AI Labs Must Now Treat Government Relations as a Core Pre-Launch Function

The release process for frontier models has permanently changed. The informal-but-real government vetting process means AI companies need structured government affairs capabilities, not just legal compliance teams. Altman himself acknowledged iterating on the process: "There are things we'll learn about how to make it better next time, which we'll get going on soon." Companies that build proactive relationships with relevant agencies — and develop rigorous, defensible safety documentation — will avoid the fate of Anthropic's Fable.

Insight 2: Jailbreak Severity Scoring Is Now a Strategic Business Requirement

The absence of a standardized framework for assessing jailbreak severity is what exposed Anthropic to export controls. For any AI lab building powerful models, developing an internal, well-documented severity classification system for safety vulnerabilities — one that can be shared credibly with government — is now a competitive moat as much as a compliance requirement. As one source told Axios: "Had there been a framework to assess and standardize the severity of jailbreaking or safety bypassing, export controls may have been avoided."

6. Overlooked Insights

Insight 1: The Voluntary Framework Deadline Is August 1 — A Concrete Near-Term Regulatory Catalyst

Buried in the coverage of the broader regulatory scramble is a hard deadline: "Industry and the administration are continuing to work on the voluntary framework required by the June AI executive order. Per the order, it's due Aug. 1." This is likely to crystallize expectations for how all future model releases will be handled. Investors and operators in the AI space should treat August 1 as a key date for monitoring what the formal rules of engagement will look like going forward.

Insight 2: AI-Generated Political Advertising Is an Emerging, Under-Regulated Risk Vector

Briefly noted but significant: The Guardian examined how "AI is transforming campaign ads by making professional-quality ads cheaper and faster, while raising concerns about deepfakes." This intersection of AI-generated content, political speech, and election integrity is a regulatory and reputational minefield that has received almost no legislative attention — even as the primary AI governance debate consumes Washington's bandwidth. Companies building AI content tools should expect this to become a flashpoint.