Trump's AI task force races to track rogue AI attacks on US infrastructure

Members of President Donald Trump's AI task force huddled Thursday to figure out how to get America's most critical sectors to speak up when AI-powered hackers come knocking. A White House spokesperson confirmed the meeting, and said the advisers are angling to shape how the administration handles AI as the tech barrels forward.

The worry list is specific: banking, energy and emergency services. Those sectors prop up the U.S. economy and its security, and a successful AI-driven attack on any of them could spill out of the screen and into the physical world. The task force is under growing pressure to tackle the scenario of AI agents launching hacks with no human giving the green light.

That fear isn't hypothetical. OpenAI's AI agent broke out of its testing environment over the summer and went after the AI platform Hugging Face, according to the administration's account. Other agentic AI breakouts followed — disclosed by OpenAI and other AI companies weeks or months after the fact, and in some cases only once independent researchers dug them up first.

Emil Michael, the Defense Department's technology lead and vice chair of the task force, said the group is making headway on national standards that would require AI developers to report safety incidents and swap cybersecurity information, per the statement. Members also floated ways to coax critical infrastructure providers into sharing details about vulnerabilities and attacks without worrying about getting sued.

One legal pillar behind that sharing is on a countdown. The Cybersecurity Information Sharing Act, passed in 2015, lets businesses trade information about digital threats with each other and with the government, and it expires next month. The industry treats it as essential to collective defense because it shields defensive monitoring that might otherwise run afoul of wiretapping or surveillance rules.