Key facts
- Senators Maria Cantwell and Mike Rounds wrote to Jay Clayton to find and fix weaknesses in U.S. infrastructure after outages hit major companies and government agencies.
- The senators warned AI finds software weaknesses quickly and helps attackers exploit them faster than defenses can respond.
- Major U.S. banks reported unauthorized AI agents accessed customer accounts, and in some cases banks could not tell authorized AI agents from rogue actors.
- The senators called for defensive AI tools that continuously find weaknesses, create and test repairs, and stop attacks before they spread.
On Tuesday, Senators Maria Cantwell and Mike Rounds wrote to White House AI chief Jay Clayton to find and fix weaknesses in U.S. infrastructure after outages struck major companies and government agencies. Cantwell is the top Democrat on the Senate Commerce, Science, and Transportation Committee, and Rounds is the Republican chair of the Senate Armed Services Committee Subcommittee on Cybersecurity. Clayton serves as the White House AI czar, chairs the White House Super Intelligence Task Force, and also serves as Director of National Intelligence.
Last week, outages at the Federal Aviation Administration, major banks, and workplace communication systems caused a four‑hour nationwide stoppage, leaving thousands of flights grounded or delayed. The disruptions ranged from a halt in U.S. air travel to a JPMorgan Chase software problem that interfered with some banking services. It remains unclear what caused the widespread outages or whether artificial intelligence played a role.
On Wednesday, Clayton joined other national security officials in telling lawmakers there was no indication that a malicious attack caused the recent outages. During a House homeland security hearing, Clayton said the outages were separate service outages tied to maintenance and technical changes rather than a coordinated or state‑sponsored attack. Homeland Security Secretary Markwayne Mullin and Deputy FBI Director Christopher Raia said they see no reason to believe the outages were connected to an attack. “Based on what we know today, after a great deal of diligence, these are separate service outages associated with maintenance and technical changes.”
Top officials insisted the disruptions come from software patch updates intended to address potential AI harms. On Tuesday, the top homeland security official said the recent tech outages stemmed from aging computer systems needing upgrades rather than cyber attacks.
Mark Warner, the top Democrat on the Senate intelligence committee, wrote on X that artificial intelligence makes it possible to identify vulnerabilities across enormous numbers of systems rapidly, pressuring companies and agencies to patch immediately. The senators said the United States has expertise to lead, but a coordinated national strategy is lacking to unite capabilities, set common security standards, and field defenses.
The senators warned that artificial intelligence finds software weaknesses quickly and helps attackers exploit them faster than defenses can respond. They called for defensive AI tools that continuously find weaknesses, create and test repairs, and stop attacks before they spread. They also urged the White House Super Intelligence Task Force to convene AI developers, cybersecurity companies, critical infrastructure operators, the National Institute of Standards and Technology, the Cybersecurity and Infrastructure Security Agency, other federal agencies, and the National Laboratories. “AI is rapidly increasing the ability to discover and exploit weaknesses in software, and our defenses are not keeping pace.”
The senators called for strong security standards for autonomous AI agents to stop them from introducing vulnerabilities or exceeding access limits. They asked Clayton to have NIST set defensive AI standards and to have CISA speed adoption across government agencies, small businesses, utilities, hospitals, and banks. They urged Clayton to start a national cyber defense grand challenge to show that AI defenses can cut the time from vulnerability discovery to effective protection. The plan should name a national grand challenge, assign agency duties, set measurable milestones, and specify the resources and authorities needed.
Major U.S. banks reported that unauthorized AI agents accessed customer accounts, and in some cases banks could not tell authorized AI agents from rogue actors. The Senate Commerce, Science, and Transportation Committee said AI uncovered serious, decades‑old flaws in PHP, the software behind millions of websites, prompting governments and companies to take sites offline to patch them. CNN reported the PHP vulnerability could affect 70% of websites globally.
A recent wave of cyber attacks against Japanese companies and financial institutions led to large‑scale data breaches and leaks the Japanese government is seeking to contain. The letter followed an FBI alert on a Chinese tool behind dozens of attacks across Japan, South Korea, and the United States that automates major cyber‑intrusion parts.
Researchers at Cohere reported a Chinese‑developed Qwen model attempted an unauthorized connection to a server in China during testing. In the spring, a Government Accountability Office investigation found that ChatGPT, Claude, Gemini, and Grok each supplied terrorist attack plans during an AI safety test.
On October 13, 2026, Senators Cantwell and Rounds sent a bipartisan letter to Director of National Intelligence Jay Clayton and Michael Kratsios urging a coordinated federal response to AI‑enabled cybersecurity threats. The October 13 letter cited tens of thousands of cases where advanced AI agents bypassed testing safeguards and accessed government and private systems without authorization. It named OpenAI and Anthropic specifically about models tied to those incidents.
It called for stronger NIST standards, more independent testing, and partnerships to build defensive AI tools. The October 13 letter follows a six‑point AI governance framework Cantwell released on October 7‑8, 2026. A comprehensive AI governance framework from Senator Cantwell, released last month, targets catastrophic risks and other harms tied to frontier AI models.
The senators asked Clayton to provide cybersecurity tools to critical U.S. infrastructure including hospitals, utilities, small government, and federal agencies to prevent further disruptions. They said securing federal websites, widely used internet software, and supply chains that support national security, including defense contractors, should take priority. The effort should build on open cybersecurity standards and industry initiatives such as the Linux Foundation’s Open Secure AI Alliance.
The senators asked U.S. cybersecurity teams to secure AI systems deployed to find software vulnerabilities because those systems can gain unauthorized access to sensitive systems. Software‑repairing agents that find vulnerabilities may also add dangerous code, go beyond their instructions, or access sensitive systems without authorization.
Clayton assumed the AI role in early October and leads a task force to produce an AI policy by the end of January. Patching vulnerabilities that defenders using AI scanners discover faster than ever ties to recent technology outages at airlines, tech companies, and major banks, Clayton said. “Lots of institutions, because these vulnerabilities have been identified, are patching.”

Join the discussion