🤖 AI arms control | AI+ Government Summary

1. Key Themes

AI safety is becoming a Cold War-style arms control issue

Rep. Ro Khanna is framing AI risk with China explicitly in nuclear-treaty terms, proposing binding agreements, inspection regimes and mutual "kill switches" rather than unilateral U.S. dominance.

"Just like America and the USSR set aside our vast differences to save the world from nuclear war, the U.S. and China share a common interest in ensuring humans never lose control of AI."

Federal AI regulation is heating up from multiple directions simultaneously

State AGs, Senate Democrats, and even a bipartisan super PAC are all pushing for stronger federal oversight — even as the White House pursues a deregulatory, export-driven strategy.

"Attorneys general from 23 states, along with D.C. and American Samoa, sent congressional leaders a letter calling for 'comprehensive federal regulation and safety protocols' for AI." Sens. Bennet and Welch "announced the AI Regulator Act, which would create a federal regulator with power to pre-certify advanced models and pause some releases for up to six months."

AI is becoming a defining 2028 political issue, with money following influence

A pro-AI industry super PAC is actively shaping the Democratic field ahead of the midterms, while facing internal party backlash — showing how deeply tech money is now embedded in electoral politics.

"Leading the Future... is backed by tech executives and investors and is spending big on AI messaging ahead of the midterms." "More than 170 Democratic organizers and campaign staff members this month called on party leaders to refuse Leading the Future money."

The Trump administration's AI export strategy is central to its global posture, despite stumbles

Rather than pursuing multilateral safety regulation, the White House is doubling down on exporting American AI as an alternative to Chinese influence.

"The U.S. is the only technology superpower willing and able to truly empower partner nations in your pursuit of a sovereign superintelligence capability." (Notably, this comes despite the program's documented struggles, referenced via the "struggling American AI Exports Program" link.)

2. Contrarian Perspectives

A Democrat is out-hawking Trump on China AI policy — via cooperation, not confrontation

While the administration signals disinterest in AI safety talks with Beijing, Khanna argues more engagement — not less — is the safer path, inverting the typical hawk/dove framing on China policy.

Trump "showing little interest in doing much" on AI with Xi, while Khanna pushes for "a binding treaty and international inspections."

Security best practice may be "control access," not "track everything"

Sponsor content (Delinea CEO Art Gilliland) pushes back on the default AI governance instinct — cataloging every agent — arguing it's fundamentally unworkable at scale.

"Most teams are trying to secure AI by inventorying every agent. It can't be done. Agents spin up faster than any list can track." The better move: "Control what agents can reach the moment they act."

3. Companies Identified

Leading the Future / Think Big PAC Pro-AI industry super PAC and affiliated political action committee. Why mentioned: First group to formally endorse Democrats for the general election, illustrating tech industry's direct electoral spending strategy amid regulatory pressure.

"The super PAC had $31 million at the end of the second quarter after transferring $20 million to affiliated groups, including $10 million to Think Big PAC."

Delinea Cybersecurity/identity management company (newsletter sponsor). Why mentioned: Advancing a specific operating philosophy on AI agent security via CEO commentary.

"Control what agents can reach the moment they act."

4. People Identified

Rep. Ro Khanna (D-Calif.) Ranking member, House China Select Committee; potential 2028 presidential contender. Why mentioned: Authored a detailed alternative U.S.-China AI policy framework sent directly to the Secretary of State.

He "calls for a global pause on recursive self-improvement... and a separate global pause on superintelligence."

Ryan Baasch White House AI adviser, deputy to NEC Director Kevin Hassett. Why mentioned: His departure signals continued churn in the administration's AI policy apparatus.

"Baasch played a critical role devising the White House's AI policy framework," including efforts "to push federal preemption of state AI laws."

Michael Kratsios White House science and technology adviser. Why mentioned: Publicly pitched U.S. AI exports as the alternative to global regulation at the UN.

"The U.S. is the only technology superpower willing and able to truly empower partner nations in your pursuit of a sovereign superintelligence capability."

Art Gilliland CEO, Delinea. Why mentioned: Offers a contrarian operating tactic for AI agent security.

"Control what agents can reach the moment they act."

5. Operating Insights

Don't build AI security around inventory/tracking — build around runtime access control. As agent creation scales, static inventories become obsolete instantly; the more durable control point is restricting what an agent can reach at the moment of action, not maintaining a registry of what exists.
Political/regulatory risk is now bifurcated and fast-moving: companies operating in AI need to track state-level coalitions (23 AGs), federal legislative proposals (AI Regulator Act), executive action (California's "kill switch" order), and electoral spending (super PACs) simultaneously — no single front determines the regulatory outcome.
Personnel turnover signals policy instability: the departure of a key architect of White House AI policy (Baasch) following "numerous others" suggests founders/operators should expect continued flux in federal AI policy positioning rather than a stable regulatory baseline.

6. Overlooked Insights

California's executive order already includes an independently verified "kill switch" mandate — this state-level action may be a preview of what federal proposals (like Khanna's or the AI Regulator Act) eventually formalize nationally, making California a bellwether for binding AI safety mechanisms rather than just a symbolic actor.

Newsom "tapped experts to advise on his recent executive order on advanced AI safeguards, including an independently verified 'kill switch' for emergencies."

A scheduled Sept. 29 meeting between Trump, Speaker Johnson, and tech CEOs was reported as a scoop with minimal detail — but its existence suggests active, direct industry-White House coordination on AI policy that operators should watch for follow-on announcements.

"President Trump, House Speaker Mike Johnson and tech CEOs will hold an AI meeting on Sept. 29, a source familiar told Maria."