AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › Source Summaries

Dominating AI Requires Understanding AI (Frazier + Rozenshtein, Lawfare, May 12 2026)

medium confidence · updated 2026-06-06

Op-ed arguing the Trump administration has pivoted from 'global dominance via limited intervention' to 'dominance by understanding AI,' anchored by a planned EO on AI lab–government cybersecurity information-sharing; proposes a five-part policy menu including CREATE AI Act + AI Talent Act, NIST funding, DPA Section 705/708 invocation, CISA workforce restoration, and a quarterly DPA frontier-lab survey.

Contents

A May 12, 2026 Lawfare op-ed by Kevin Frazier and Alan Z. Rozenshtein arguing that the Trump administration has moved from a posture of global AI dominance through limited intervention toward what the authors call "dominating AI by understanding it." The piece, published in Lawfare's Cybersecurity & Tech / Executive Branch section (Lawfare), proposes a policy menu spanning congressional and executive-branch action, built around invocation of the Defense Production Act (DPA) and restoration of the federal cybersecurity workforce.

Central argument

The authors characterize the prior debate as "doomers vs. accelerationists" — predeployment government testing on one side, unfettered global dominance via limited intervention on the other — and argue a third Trump-administration posture has emerged: dominating AI by understanding it. They read the shift as signaled by a planned executive order that would facilitate greater information-sharing between AI labs and government cybersecurity programs, paired with renewed senior-White-House engagement with frontier-lab CEOs in the weeks before the article (which they attribute to WSJ reporting on a cluster of Trump–Anthropic–Mythos meetings).

The authors ground the pivot in three rationales:

  1. Cybersecurity and biosecurity risks of frontier AI are being taken seriously, citing Economist coverage of bioterrorism uplift and recent CISA warnings about destructive cyberattacks.
  2. Adoption-lag awareness: military and private-sector deployment of the most powerful models will stall until reliability, effectiveness, and controllability are established.
  3. Recognition that technical AI progress (understanding the model) and dominance (deploying the model at scale) are connected rather than opposed.

Proposed policy menu

The article frames its recommendations as a menu split between Congress and the executive branch.

For Congress, the authors call to pass the CREATE AI Act (Young) (NAIRR codification plus a funding increase); pass the AI Talent Act (H.R. 6573), a federal AI-workforce expansion; and appropriate more funds to NIST, specifically the Center for AI Standards and Innovation (NIST CAISI (Center for AI Standards and Innovation)). In support, the article cites a Center for a New American Security (CNAS) gap finding that "for every dollar allocated to NAIRR, the private sector is investing roughly $23,000 in AI." It notes that National Security Commission on Artificial Intelligence (NSCAI) called for $32B in non-defense AI R&D by FY2026, and that realized investment is "about a tenth of that" per Center for Strategic and International Studies (CSIS) analysis.

For the executive branch, the authors propose three actions:

  • DPA Section 708 voluntary agreements to coordinate frontier labs on cyber- and bio-security mitigation methods and conditional non-deployment agreements, under an Attorney General plus FTC chair antitrust-defense screen. The authors argue the existing Frontier Model Forum can support best-practices coordination but cannot lawfully coordinate release decisions without Section 708 antitrust cover.
  • CISA workforce restoration. The article states the administration previously cut roughly 1,000 employees (about a third of the CISA workforce), that the May 2026 75-day DHS shutdown furloughed the remainder, and that a plan to hire 300 mission-critical positions is, in the authors' words, "a surge in capacity [that] could not come at a better time." They push for additional hiring beyond the 300, alongside use of the Cyber Response and Recovery Fund (6 USC § 677a) for critical-infrastructure entity support.
  • A DPA Section 705 quarterly frontier-lab capability survey, confidential and including internal-only deployed models. The article cites a capability-disclosure-gap finding (arXiv 2504.12170) to argue that labs do not always disclose internal-deployment capabilities publicly, leaving the government without visibility.

Relation to other policy frames

The authors present "dominance by understanding" as a third pole distinct from "doomer" pre-release testing mandates (which the authors note were rejected by the post-Mythos cyber-governance bidding cluster — see Analysis: Lawmakers, industry pitch frontier AI governance approaches as they await White House moves (Inside AI Policy, May 15 2026)) and from "accelerationist" minimal-intervention dominance (as set out in America's AI Action Plan, July 2025).

The frame is related to but distinct from the Responsible Scaling Policy (RSP) / Anthropic's Responsible Scaling Policy (Version 2.2) idea: where responsible scaling policies are voluntary lab commitments to capability-dependent gating, "dominance by understanding" is described as a government-coordinated R&D investment plus DPA-leverage stance. The CISA-staffing recommendation parallels Cyberwar's New Frontier: How AI Agents Will Threaten Global Security (Rosen + Kraprayoon, Foreign Affairs, April 16 2026), which advances a comparable call to restore pre-2025 cybersecurity-workforce levels.

Relationships

Provenance

  • Raw Sources/Dominating AI Requires Understanding AI.md — full text
  • Published at Lawfare: lawfaremedia.org
Categories: Source Summaries | us | opinion | frontier-models | safety | national-security | government | executive-order | framework
What links here (7)