Urgent.News

What's breaking now, across thousands of outlets.

AI

The US wants an AI-era 'red phone' with China

President Donald Trump and Chinese leader Xi Jinping are expected to discuss AI safety and trade at their summit in Washington this week.

Treasury Secretary Scott Bessent announced a US proposal for an AI-based direct line of communication with China, amid escalating tensions in the rivalry between the two nations. The aim is to provide a notification mechanism for major AI-related national security incidents, as Washington seeks transparency in the burgeoning AI safety debate.

The discussion took place during talks in New York between Bessent and Chinese Vice Premier He Lifeng, preceding President Donald Trump's summit with Chinese leader Xi Jinping in Washington. Bessent emphasized the importance of transparency between the world's leading AI powers, stating that moving from opacity to more transparency is crucial for both countries.

Written by urgent.news from Business Insider's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at businessinsider.com →

More in AI

I built the same agent in Strands, LangGraph, and CrewAI — and recorded every LLM call to see how they actually differ

Which framework should I actually use? Every comparison article has an opinion. Almost none of them has data. So I built the same agent three times — once in Strands , once in LangGraph , once in…

  • Strands framework uses model-driven approach with adaptive LLM calls
  • LangGraph employs graph-driven approach with deterministic workflow
  • CrewAI takes role-based framework with consistent output

When A Legitimate Sounding Request Quietly Breaks The Bot's Actual Scope

Most conversations about keeping a custom bot inside its intended boundaries focus on obvious misuse, someone deliberately trying to manipulate it into ignoring its instructions.

  • Users make legitimate requests that exceed bot's capabilities
  • Model complies without realizing scope boundaries are crossed
  • Issue grows as users extend requests, assuming capability persists

More from Monday 21 September →