Industry cybersecurity leaders say the White House’s new voluntary accord with leading artificial intelligence (AI) companies puts greater emphasis on oversight and accountability, but federal agencies will also need to prioritize operational resilience as AI systems become more powerful.
President Donald Trump and executives from leading AI companies signed the “White House Accord on Super Intelligence” on Sept. 29, establishing a voluntary framework for companies to monitor and audit advanced AI systems. Trump called the agreement “morally binding” and said participating companies would help police one another.
Gary Barlet, public sector chief technology officer at cybersecurity company Illumio, said the agreement points in the right direction by emphasizing accountability as AI capabilities advance.
“The accord gets one important thing right: As AI capabilities advance, security and accountability can’t be optional,” Barlet told MeriTalk. “Greater transparency, testing, and shared responsibility are necessary for organizations to deploy increasingly powerful AI systems at scale.”
However, Barlet said the framework should put greater focus on what happens after an AI system is deployed and operating inside an agency environment.
“What may be missing is enough focus on operational resilience,” Barlet said. “Most AI governance discussions center on the models themselves, but agencies also need to consider the environments in which those systems operate, the data they can access, and what happens when something goes wrong. AI doesn’t eliminate risk. In many cases, it expands the attack surface.”
“For agencies, the priority should be building the visibility and controls needed to understand where AI is being used, what it can access, and how to contain the impact of a failure or compromise,” he continued. “The question isn’t whether AI systems will make mistakes. The question is whether agencies are prepared when they do.”
Bill Wright, head of government affairs at storage and data management company Everpure, similarly pointed to the accord’s governance structure as a potentially useful model.
“The structure laid out in the morally binding accord mirrors approaches organizations already use to manage financial and cybersecurity risk: establish internal controls, evaluate them, and make sure outside accountability reaches the board,” Wright told MeriTalk.
“For that oversight to truly work, auditors need access to the underlying data and records, including what data was used to train AI models, how those models changed over time, and what they did once deployed,” he said.
Wright also cautioned agencies against treating audit records and other underlying AI data as separate from their broader cyber resilience responsibilities.
“Agencies also need to assume something will go wrong,” Wright said. “If an attacker can encrypt or alter the data and logs behind an AI system, the agency can lose both the system and the audit trail. Locked-down copies and tested recovery processes need to be part of AI governance.”
“Additionally, agencies need consistent controls across environments and the ability to move their data as models and vendors change,” he added.
The debate about oversight continues as the Trump administration is also changing how the federal government refers to the technology.
Trump signed a Sept. 29 executive order directing executive branch agencies to replace the terms artificial intelligence and AI with super intelligence and SI in official communications, websites, reports, policy documents, and other non-statutory materials.
The order currently defines SI using the existing statutory definition of AI. It also gives the assistant to the president for science and technology 60 days to submit proposed legislative language for a federal SI definition and assess whether existing statutory references to AI should be changed.
For Barlet, the significance of that order will ultimately depend less on the terminology than on whether the policies surrounding it can withstand changes in the technology.
“The long-term value of the SI executive order won’t depend on whether it accurately predicts AI’s future,” Barlet said. “It will depend on whether it establishes durable principles around security, accountability, and resilience that can adapt as the technology evolves.”
“The agencies that succeed won’t be the ones trying to predict every AI risk,” he added. “They’ll be the ones building resilient operations that can withstand surprises.”
Wright also said the terminology change does not alter the fundamental technology-management challenges facing agencies.
He observed, “whether we call it artificial intelligence or super intelligence, the underlying work remains the same: keep the data secure and accessible, know who is accountable, and maintain control as technologies, models, and vendors evolve.”