White House Request Raises AI Safety Center Independence Concerns

By Olivia Hart

CAISSI, the Commerce Department’s Center for Advancing Innovation and Standards for Super Intelligence, has a $10 million annual allocation and no permanent director, AP reported. On November 8, 2026, AP reported that White House officials had asked CAISSI to report informally to a White House office; no formal reorganization was announced.

Who now directs CAISSI?

Two former government officials familiar with the discussions described the request on condition of anonymity to protect current staffers. AP did not identify the office and called the arrangement informal rather than an announced reorganization.

CAISSI remains an independent Commerce Department division, formerly the U.S. Center for AI Standards and Innovation, AP reported. Its $10 million allocation is about one-eighth its United Kingdom counterpart’s annual budget.

Its director was appointed in April 2026 and fired within days. A successor resigned after a few months; the head of its parent institute now serves as acting director. The center has formal standing but unsettled leadership.

Why does the reporting line matter?

AP reported that critics feared the arrangement could politicize cybersecurity and bioweapon decisions. They warned findings might be altered or muted to accommodate an industry largely entrusted to police itself.

Caroline Siegel Singh, an AI and emerging technology program manager at the Federation of American Scientists, argued for nonpartisan or bipartisan technocratic experts to make policy in the public interest. Others told AP that closer ties to President Donald Trump could give the division greater prominence and possibly more funding. Proximity offers visibility and money; independence protects credibility.

What is CAISSI supposed to evaluate?

NIST calls CAISSI industry’s primary government contact for testing and collaborative research on commercial super-intelligence systems. Its mandate includes:

• Developing measurement guidelines, best practices and voluntary standards.
• Establishing voluntary agreements with developers and evaluators.
• Leading unclassified evaluations of demonstrable national-security risks involving cybersecurity, biosecurity and chemical weapons.
• Assessing adversary systems for vulnerabilities, backdoors and other covert malicious behavior.
• Coordinating evaluation methods and assessments with the Defense and Energy departments, the Department of Homeland Security, the Office of Science and Technology Policy, and the Intelligence Community.

Voluntary standards sit alongside government-led evaluations. Technical independence matters because the reporting request could influence which risks receive attention.

Can Washington block a model release?

Executive Order 14409, issued June 2, 2026, directs national-security agencies, consulting NIST and others, to develop a classified process for benchmarking advanced cyber capabilities. It would establish the threshold for a “covered frontier model.”

The order calls for voluntary government access to covered frontier models for up to 30 days before their planned release to other trusted partners. It does not authorize mandatory licensing, preclearance or permitting for developing, publishing, releasing or distributing AI models.

As AP framed it, the dispute concerns who directs and communicates technical findings. Washington still lacks a new power to approve models before release.

Can private reviews replace public testing?

In September 2026, President Donald Trump asked leading AI companies to undertake extensive self-policing through a voluntary safety pact. AP reported that the administration has not explained how it will monitor implementation.

AP said the evaluation problem changes as models become AI agents that plan and execute multi-step tasks without continual human prompting. Sayash Kapoor, an incoming assistant professor at the University of California, Berkeley School of Information, said no standard process exists for testing agents’ safety in real-world settings.

Kapoor said an agent booking a flight can appear successful if only the final result is checked. “But unless you’re looking at the entire set of steps that an agent took, you won’t be able to tell that the agent made a huge mistake,” he said.

Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have promoted independent evaluation firms. Their companies can restrict access and set aside evaluations, AP reported.

Logan Koepke, senior project director at the civil rights and technology nonprofit Upturn, said company-run internal evaluations and company-hired external reviewers leave government testing’s role unclear. Amos Toh, senior counsel at the Brennan Center for Justice, said letting laboratories police their own models without clarifying government’s role weakens independent checks.

Toh said consequential decisions should be explained and justified publicly. Independent reviewers do not solve the problem when model makers control access and can set aside evaluations.

What remains unresolved?

AP reported that Congress began asking in August 2026 why CAISSI had stopped releasing reports on how safely certain AI systems were performing. OpenAI also urged policymakers to provide the division with more funding and support.

NIST’s official CAISSI page lists a March 23, 2026 research-blog entry as its most recent. Its news section includes a September 17, 2026 assessment of Z.ai’s GLM-5.3 cyber capabilities and an October 16, 2026 publication on incident readiness and response. Those entries show continuing work, but do not answer Congress’s question.

The White House did not respond to AP’s specific questions about CAISSI. It said Trump had formally established a Super Intelligence Force and commissioned a risk report from the Council of Advisors on Science and Technology.

The unresolved question is who sets CAISSI’s priorities and how visible its methods and findings will be. Closer White House ties may add prominence, but AP said they do not resolve independent government testing alongside voluntary developer access and company-controlled reviews. CAISSI’s credibility will remain unsettled until priorities, methods and disclosures are clear.