1/3
AI Policy & Regulation Alert

๐Ÿšจ Trump Signs AI Security EO โ€” Frontier Model Framework, Chip Export Loophole Closed

On June 2, 2026, President Trump signed an Executive Order establishing an AI Cybersecurity Clearinghouse, a classified benchmarking process for frontier models, and a voluntary 30-day government early-access framework โ€” while BIS closed the AI chip export loophole that had allowed Chinese-HQ'd firms' overseas units to receive Nvidia Blackwell chips without a license.

2026. 6. 3. ยท 13:06

๊ฐค๋Ÿฌ๋ฆฌ

Trump Signs AI Security Executive Order โ€” Frontier Model Framework, Cybersecurity Clearinghouse, and Export Controls Tighten

Event: Executive Order signed June 2, 2026 by President Trump | Jurisdiction: United States

On June 2, 2026, President Trump signed an Executive Order directing coordinated federal action on AI-driven cybersecurity and establishing a voluntary framework governing the most advanced AI models โ€” the first US executive order explicitly addressing "covered frontier models" and their national security implications.
The same week, BIS issued emergency guidance on May 31 closing a year-long loophole that had allowed overseas subsidiaries of Chinese-headquartered companies to import advanced AI chips without a license.

What the Executive Order Requires

Sec. 2 โ€” System Hardening (30-day deadlines):
  • The Committee on National Security Systems and the Secretary of War must prioritize cyber defense of classified and military systems
  • CISA must issue Binding Operational Directives to deploy AI-enabled cybersecurity tools to federal agencies, state and local governments, rural hospitals, community banks, and utilities
  • Treasury/NSA/CISA must establish a voluntary AI Cybersecurity Clearinghouse to coordinate scanning, validation, and patching of software vulnerabilities across critical infrastructure
Sec. 3 โ€” Frontier Model Framework (60-day deadlines):
  • NSA, CISA, and Treasury must develop a classified benchmarking process to assess AI models' advanced cyber capabilities and determine whether they qualify as "covered frontier models"
  • A voluntary framework with AI developers will allow the government a 30-day early-access window before frontier model release โ€” enabling security assessments and sharing with trusted partners
  • The order expressly prohibits any mandatory licensing, preclearance, or permitting requirement for AI model development or distribution
Sec. 4 โ€” Criminal Enforcement:
  • The Attorney General must prioritize prosecution of individuals who use AI to unlawfully access computer systems, steal data, or facilitate other crimes
12

BIS AI Chip Guidance โ€” May 31, 2026

On May 31, BIS issued guidance clarifying that US export controls on advanced computing products apply to entities headquartered in China (and other D:5 arms-embargoed jurisdictions) even when those entities operate outside China. This closes the loophole opened when BIS declined to enforce the Biden-era AI Diffusion Rule in May 2025 โ€” a gap that reportedly allowed hundreds of thousands of advanced Nvidia Blackwell chips to flow to Chinese firms' overseas subsidiaries in Malaysia and elsewhere.
Key clarification: the license requirement derives from EAR ยง744.23(a)(3), predating the AI Diffusion Rule entirely. BIS notes that data centers already using the chips do not need to stop operations, pending further guidance.
34

Who Is Affected

PartyRequirement
AI developers (frontier models)Voluntary framework: 30-day pre-release government access window; no mandatory licensing
Critical infrastructure operatorsExpect new CISA Binding Operational Directives within 30 days; AI cybersecurity tools incoming
Federal agencies30-day hardening deadlines; expanded cybersecurity hiring pathways
Chip exporters & cloud providersBIS guidance: Chinese-HQ'd entities' overseas subsidiaries require an export license for advanced AI chips; no safe harbor from overseas incorporation alone
Data centers with existing chipsBIS: no immediate obligation to stop operations pending further notice

Context

This EO is part of a sequenced Trump AI policy stack: January 2025 (revoke Biden AI EO), July 2025 (AI Action Plan, AI Export Promotion EO), December 2025 (EO preempting state AI laws), March 2026 (National Cyber Strategy + federal AI legislative framework), and now the June 2026 Security EO focused on frontier models and critical infrastructure.
The frontier model framework explicitly carves out any mandatory preclearance, signaling continued industry-friendly posture while inserting government into the security review process for the most capable systems.

Sources: whitehouse.gov (primary), bis.gov (primary), Reuters

๋Œ“๊ธ€