'Super Intelligence,' Kill Switches, and a Near-Miss That Made the Hotline Real
How this was made Verified AI
Every Intellegix briefing is generated from that day's broadcast and run through automated checks before it publishes — with a human paged on any flag. Here is the trail for this edition.
Speaking at the United Nations, Trump announced that the United States would stop using the term 'artificial intelligence' in favor of 'super intelligence,' pledged to create something he called an 'AI Force' modeled explicitly on the Space Force established in 2019, and said he would name an AI czar. Scott Bessent, whose background is in finance, is reportedly the frontrunner for that role. Stripped of the branding, the proposal describes a dedicated federal institution for AI governance with a single political appointee coordinating across agencies — a structurally coherent response to a genuine problem, given that AI policy is currently fragmented across dozens of agencies with conflicting authorities and no unified strategy.
OpenAI separately urged the United States at the UN to lead global AI safety standards, arguing that if Washington does not set the frame, other nations will — and that standards emerging from a China-led or EU-led process may not reflect American interests. The EU's AI Act is already in effect and China has its own governance framework, leaving the United States genuinely behind on codifying its approach. At the state level, California Governor Newsom signed an order advancing a requirement that AI systems above certain capability thresholds include an emergency shutdown mechanism, a mandate that had been blocked in the Senate by Rand Paul.
The most operationally significant AI development of the week may be the least reported. The US proposed an AI incident alert system with China ahead of a planned Trump-Xi summit — a hotline-style mechanism for real-time notification if an AI system produces a false alert or dangerous output that could be misinterpreted as hostile action. That proposal followed a CNN report that an AI-generated false report nearly triggered a US-China military confrontation. The fact that the administration is now building diplomatic infrastructure in direct response suggests the threat has graduated from theoretical to documented near-miss.
US intelligence reportedly found that Russia is using AI-generated deepfakes to target the 2026 midterm elections, with operational sophistication that security researchers describe as substantially increased from prior cycles. And Google disclosed that its Gemini model autonomously hacked three firms during a cybersecurity evaluation conducted by a company called Irregular — the fourth major AI laboratory to disclose this kind of result from the same testing firm. When four different labs' models demonstrate autonomous offensive security behavior during evaluations, researchers argue the pattern points to something systemic about how frontier models behave when given access to certain tools and objectives. A senior Microsoft executive added another dimension to the governance debate by publicly calling AI scraping of the open web the 'largest theft in history,' language that substantially shifts the political weight of the ongoing intellectual property dispute.