The Office of the National Cyber Director (ONCD) is a White House office headed by the National Cyber Director, a post held in the second Trump administration by Sean Cairncross. From spring 2026 its director led the federal response to Anthropic's Claude Mythos Preview, and the office became one of the agencies involved in U.S. pre-release review of frontier AI models. It is a consulted office under the clearinghouse and covered-frontier-model provisions of Executive Order 14409, it asked OpenAI with the Office of Science and Technology Policy in June 2026 to limit the initial release of GPT-5.6, and its director is among the officials heading the voluntary review framework the order required. It is the lead defendant in Protect Democracy Project v. Office of the National Cyber Director, a Freedom of Information Act suit seeking that framework's text.
Overview
ONCD is a White House office (Source: axios.com; reuters.com). Sean Cairncross is the National Cyber Director in the second Trump administration (Sources: wsj.com; insideaipolicy.com; Analysis: Lawmakers, industry pitch frontier AI governance approaches as they await White House moves (Inside AI Policy, May 15 2026)). Kemba Walden is a former head of the office; on May 4, 2026 she raised the prospect of frontier-AI firms collaborating with NIST's vulnerability-disclosure program (Source: insideaipolicy.com). Will Loucks is the office's senior director for intelligence (Source: insideaipolicy.com).
The National Cyber Director is named in the principal 2026 presidential instruments on AI and national security. NSPM-11, signed June 5, 2026, is addressed to a slate of principals that includes the National Cyber Director alongside the Secretaries of State, Treasury, War, Energy, and Homeland Security, the Attorney General, the DNI and CIA Director, and the OMB Director (NSPM-11 — Artificial Intelligence in the National Security Enterprise (Trump, signed June 5, 2026)). Executive Order 14409 of June 2, 2026 assigns the director a consulting role in four of its provisions, described below (Executive Order: Promoting Advanced Artificial Intelligence Innovation and Security (signed June 2, 2026)).
Response to Claude Mythos Preview
ONCD convened roughly 30 industry representatives on April 28, 2026 to discuss an expansion of Project Glasswing (Source: insideaipolicy.com), Anthropic's initiative with partner companies to use Claude Mythos Preview for defensive cybersecurity (Project Glasswing: Securing Critical Software for the AI Era).
The Wall Street Journal reported on May 7, 2026 that Mythos had thrown the administration's hands-off AI strategy into disarray. An April phone call in which Vice President JD Vance spoke with Sam Altman, Dario Amodei, Elon Musk, Sundar Pichai, and Satya Nadella convinced Vance that systems like Mythos could autonomously launch cyberattacks against small-town banks, hospitals, and water plants. National Cyber Director Sean Cairncross was leading the response and had formally asked Anthropic to hold off on expanding Mythos access to critical-infrastructure operators, while the White House weighed an executive order to create a formal oversight process for the most advanced AI models (Sources: wsj.com; insideaipolicy.com). Reuters reported on June 5, 2026 that Anthropic had briefed Cairncross on Mythos as tensions between the company and the White House eased ahead of its IPO (Source: reuters.com).
Institutional contest over model evaluation
ONCD took part in the May 2026 dispute within the administration over which agency would lead frontier-model evaluation. The New York Times reported on May 4, 2026 that the working group then under consideration would likely examine a formal government review process for new AI models, with oversight potentially shared across multiple agencies, including the NSA, the White House Office of the National Cyber Director, and the Director of National Intelligence, and a possible role for CAISI (Source: nytimes.com).
According to a May 11, 2026 Washington Post report, ONCD proposed creating a large center within the Office of the Director of National Intelligence (ODNI) to evaluate new AI models, which would give intelligence agencies a significant new role in AI policy. The Department of Commerce, home of CAISI, strongly opposed the idea (Analysis: Lawmakers, industry pitch frontier AI governance approaches as they await White House moves (Inside AI Policy, May 15 2026)). On May 11 the Commerce Department deleted from its website its May 5 announcement that CAISI had signed pre-deployment evaluation agreements with Microsoft, Google, and xAI; the deletion reportedly came at the urging of ONCD. Reporting framed the deletion as a sign that ONCD preferred a different institutional home for pre-release evaluation, with some accounts pointing to an ODNI-led center, and whether it reflected one-off political housekeeping or a broader withdrawal of CAISI from the pre-deployment-vetting role was not clear (Source: reuters.com). Inside AI Policy's Charlie Mitchell framed the dispute as one over which institution owns frontier-model evaluation, with ODNI and the intelligence community, Commerce and CAISI, and Treasury each seeking the lead role (Analysis: Lawmakers, industry pitch frontier AI governance approaches as they await White House moves (Inside AI Policy, May 15 2026)).
After Executive Order 14409 was signed, the White House, through Cairncross and with Treasury Secretary Scott Bessent, directed CAISI to halt publication of its model assessments while the order is implemented. The pause, reported June 9, 2026, was described as reflecting a push by Cairncross and Bessent to give national-security considerations a larger role in how AI models are evaluated, and it put the unit's public-facing future in doubt despite praise for its work from AI developers (Source: wsj.com).
Executive Order 14409
The predecisional May 2026 draft of the cybersecurity-and-frontier-AI order directed Treasury, the NSA, and CISA, with NIST, the National Cyber Director, and the Assistant to the President for Science and Technology (APST), to develop a classified benchmarking process for designating "covered frontier models" and a voluntary framework giving the government up to 90 days of pre-release access to such models (Draft Executive Order: Promoting Advanced Artificial Intelligence Innovation and Security (unsigned, May 2026)). When the signing was postponed on May 21, 2026, ONCD signaled that it was working on separate AI security initiatives, suggesting the cyber-governance agenda might proceed outside the order (Source: axios.com).
The order as signed on June 2, 2026, later identified as Executive Order 14409 (Source: politico.com), names the National Cyber Director in four provisions (Executive Order: Promoting Advanced Artificial Intelligence Innovation and Security (signed June 2, 2026)):
| Provision | Lead | ONCD role |
|---|---|---|
| Sec. 2 — Binding Operational Directives to expedite cyber defense of civilian federal systems and facilitate access to cybersecurity tools, "including where appropriate covered frontier models" | DHS, through CISA | Consulted, with OMB and the APNSA |
| Sec. 2 — AI cybersecurity clearinghouse to coordinate vulnerability scanning, validation, and patch distribution | Treasury | Consulted, with the NSA and CISA |
| Sec. 2 — 30-day determination of whether federal grant programs can fund advanced AI vulnerability detection | OMB | Consulted, with CISA |
| Sec. 3 — classified benchmarking process for advanced cyber capabilities and voluntary framework for up to 30 days of pre-release government access | Treasury, NSA, and CISA | Consulted, with the APST and Commerce through NIST; the NSA Director designates covered frontier models in consultation with ONCD, the APST, CISA, and Department of War representatives |
Section 3(c) bars construing the section to authorize "a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models" (Executive Order: Promoting Advanced Artificial Intelligence Innovation and Security (signed June 2, 2026)).
The voluntary review framework
The framework required by the order was declared complete on August 3, 2026 without publication; a White House official declined to say what it contains, who has seen it, or when companies would begin using it (Source: axios.com). Reporting the same day said that no person or organization had been designated to lead outreach with the AI companies, and named Cairncross, Bessent, and Commerce Secretary Howard Lutnick among those heading the initiative (Source: cnn.com).
The Gold Eagle clearinghouse
The clearinghouse launched on July 14, 2026 as the "Gold Eagle" initiative, a federal clearinghouse for AI-discovered cybersecurity vulnerabilities that uses frontier AI models, including Anthropic's Mythos, to help federal agencies, critical-infrastructure operators, and AI developers find and patch software flaws. It was the first program implementing the order and arrived more than a week past its July 2 deadline, with Treasury, DHS/CISA, and the Department of War participating (Source: whitehouse.gov; politico.com). Cairncross credited open-source AI developers at the launch, amid reports of a possible follow-on executive order addressing open-source AI security risks (Source: politico.com; theinformation.com). In remarks published July 20, 2026, ONCD's Will Loucks highlighted the clearinghouse as a key cybersecurity benefit of the order (Source: insideaipolicy.com).
Release request for GPT-5.6
On June 25, 2026, ONCD and the Office of Science and Technology Policy asked OpenAI to limit the initial release of GPT-5.6 to government-approved partners, reported as the first time the U.S. government had preemptively asked an American lab to restrict a launch before release. A source told Axios the government intervened because GPT-5.6 has "Mythos-like" capability, rather than as a general shift to a heavier hand (Source: axios.com). Sam Altman told staff the government would be "approving access customer by customer" during a preview period (Source: axios.com; techcrunch.com). OpenAI said on June 26, 2026 that "at their request" it was restricting initial access to a small group of trusted partners whose participation was shared with the government, a process it said "should not become the long-term default" (Source: openai.com; cybersecuritydive.com).
On July 7 the Commerce Department cleared OpenAI for broad release of GPT-5.6 after testing by CAISI, with public launch set for July 9 (Source: axios.com; cnbc.com). A White House official said no formal government approval had been required for the launch (Source: axios.com).
Request to withhold models from the UK AI Security Institute
A White House request that OpenAI and Anthropic not share new models with the UK AI Security Institute until they have been tested by the U.S. government became public on September 24, 2026; the date of the request itself was not reported. A person familiar with the matter and a senior administration official said the request came from ONCD, and the official said "this has been our policy with every new frontier model that comes out." Anthropic did not provide Claude Mythos 5.1 to UK AISI, saying the model was "only available to a set of U.S. organizations," while AISI Director Henry de Zoete told a UK parliamentary committee earlier in September that the institute had tested OpenAI's GPT-6 Astra before its release (Source: politico.com).
Chinese open-weight models
Nathan Lambert, quoting Axios reporting, lists a threat advisory considered by the NSA and ONCD among the measures the administration weighed against Chinese AI labs and models, alongside a Commerce Entity List designation, an executive order conditioning U.S. hosting of Chinese models on security guarantees and breach liability, and draft supply-chain rules. He argues these measures would leave the United States in "a very asymmetric state" in which the best U.S. models carry guardrails on cybersecurity tasks while global actors have access to Chinese open-weight models with which to probe U.S. defenses (Kimi K3: The open-weights escalation (Nathan Lambert, July 2026)).
Congressional, industry, and civil-society engagement
On May 13, 2026, 32 House members of both parties wrote to Cairncross calling for an "expansive interagency initiative" on cybersecurity and advanced AI (Analysis: Lawmakers, industry pitch frontier AI governance approaches as they await White House moves (Inside AI Policy, May 15 2026)). The letter asked ONCD to convene a federal-industry process on the disclosure of vulnerabilities found by AI systems, to expand controlled defensive access to such tools for trusted defenders, and to recommend whether and how the government can help the American software ecosystem validate, triage, and patch the resulting volume of findings, arguing that frontier systems can "discover, analyze, and exploit software vulnerabilities at a scale that existing public and private processes are not equipped to handle" (Source: latta.house.gov). It rejected mandatory pre-release review and asked ONCD to "establish a voluntary framework for AI labs to provide early access to vetted defenders" and "a process for monitoring sudden frontier AI capability jumps." Signatories included Bob Latta (R-OH), Doris Matsui (D-CA), Jay Obernolte (R-CA), Ted Lieu (D-CA), and John Moolenaar (R-MI) and Ro Khanna (D-CA), chair and ranking member of the House select committee on China (Analysis: Lawmakers, industry pitch frontier AI governance approaches as they await White House moves (Inside AI Policy, May 15 2026)). The House Homeland Security Committee received a closed-door briefing and live demonstration of Mythos Preview on May 12–14, 2026, concurrent with the letter (Source: insideaipolicy.com).
Also on May 13, an open letter to ONCD led by the Independent Community Bankers of America (ICBA) and co-signed by the Business Software Alliance and TechNet urged "sustained collaboration among frontier AI developers, relevant Federal agencies, and critical infrastructure stakeholders" on the testing, evaluation, and red-teaming of advanced models. ICBA's Anjelica Dortch called for the Treasury Department, in coordination with ONCD, CISA, and private-sector stakeholders, to engage frontier developers on voluntary safeguards against synthetic checks, fraudulent financial documents, and synthetic identities. Both letters rejected mandatory pre-release review and preferred ONCD acting as a coordinator rather than an oversight authority (Analysis: Lawmakers, industry pitch frontier AI governance approaches as they await White House moves (Inside AI Policy, May 15 2026)).
Cairncross was one of six officials addressed in an August 3, 2026 letter led by Sen. Kirsten Gillibrand and cosigned by Sens. Mark Warner, Mark Kelly, Adam Schiff, and Chris Coons. The senators argue that the administration's "ad hoc and unpredictable approach" to restricting frontier-model access undermines U.S. competitiveness, and request an unclassified response within 30 days to nine questions, among them the legal authorities invoked and the division of responsibility among the agencies named (Senate letter on the Administration's approach to limiting access to advanced AI models (Gillibrand, Warner, Kelly, Schiff, Coons, August 2026)). A September 9, 2026 letter from 28 organizations, hosted by Americans for Responsible Innovation and asking the administration to publish the review framework, was copied to Cairncross (Coalition letter to President Trump on public release of the White House frontier-AI review framework (September 2026)). An August 2026 Institute for Progress report recommends that the NSA, CAISI, CISA, and ONCD further invest in cybersecurity resilience (How Should the US Prepare for Increasingly Automated AI R&D? (IFP, August 2026)).
Litigation
In August 2026 Protect Democracy submitted identical Freedom of Information Act requests to the four agencies it later sued, seeking non-classified records about the framework the White House said on August 3, 2026 it had finalized for reviewing advanced AI models before release, the identity of participating companies, and the legal authority for the program. It requested expedited processing; only ONCD responded, and it denied expedition (Source: arstechnica.com).
On September 1, 2026 Protect Democracy sued ONCD, the Office of Science and Technology Policy, the Bureau of Industry and Security within the Department of Commerce, and the Department of the Treasury in the U.S. District Court for the District of Columbia, in Protect Democracy Project, Inc. v. Office of the National Cyber Director, No. 1:26-cv-03064. The caption abbreviates the lead defendant as "Office of the National Director," and the complaint's exhibits include ONCD's response to the request and the administrative appeal from it. The case was reassigned to Judge Ana C. Reyes on September 3 (Source: courtlistener.com; cdn.arstechnica.net). The complaint seeks production of the framework's "unclassified procedural and contractual architecture," including its text, the terms of participation, the identity of participants, and the criteria by which access to frontier models is granted or withheld (Source: protectdemocracy.org; arstechnica.com).
Protect Democracy withdrew its motion for a preliminary injunction on September 15, 2026; the court vacated the scheduled hearing and directed the parties to file a joint status report on November 13, 2026 (Source: courtlistener.com). Protect Democracy records an initial production dated September 8, 2026, and Gary Marcus, in an account published September 15, described 132 pages in which nearly all substantive content is redacted. Marcus attributes the production to OSTP, while the complaint and the organization's case page record ONCD as the only agency that answered the expedition request; which agency produced the pages is not resolved by the available sources (Source: protectdemocracy.org; garymarcus.substack.com).
Relationships
- litigates: Protect Democracy Project v. Office of the National Cyber Director — lead defendant in the FOIA suit over the frontier-model review framework
- related: Sean Cairncross — National Cyber Director heading the office
- related: AI Pre-Release Vetting — the review program in which the office is a consulted agency and its director a lead official
- related: EO — Promoting Advanced AI Innovation and Security (Trump, signed June 2, 2026) — Executive Order 14409, which names the National Cyber Director in Secs. 2 and 3
- related: NSPM-11 — Artificial Intelligence in the National Security Enterprise (Trump, June 5, 2026) — NSPM-11, addressed in part to the National Cyber Director
- related: NIST CAISI (Center for AI Standards and Innovation) — CAISI announcement deleted at ONCD's reported urging; assessment publication halted by the White House through Cairncross
- related: Office of Science and Technology Policy (OSTP) — co-requester of the GPT-5.6 release limit and co-defendant
- related: National Security Agency (NSA), Cybersecurity and Infrastructure Security Agency (CISA) — regulator role, Bureau of Industry and Security (BIS), Scott Bessent, Howard Lutnick — agencies and officials sharing roles under the executive order and the review framework
- related: Claude Mythos Preview, GPT-5.6 (Sol, Terra, Luna), Anthropic, OpenAI — models and developers subject to the office's 2026 access requests
- related: Protect Democracy — plaintiff in the FOIA suit
- related: AI and Cybersecurity