Urgent.News

What's breaking now, across thousands of outlets.

AI

China, US agree to set up channel for AI safety incidents

Following the summit between US President Donald Trump and Chinese President Xi Jinping, the two superpowers pledged to set up a "communication channel" for AI incidents and boost coordination on military matters.

Beijing and Washington have agreed to establish a communication channel for discussing AI safety incidents between their respective militaries. These developments came from the three-day summit held between US President Donald Trump and Chinese President Xi Jinping. The agreement follows a mutual tariff reduction on around $30 billion worth of non-essential goods, which includes exports such as agricultural products and holiday decorations.

The Chinese Foreign Ministry confirmed that both sides will strengthen exchanges and discussions regarding AI technologies, aiming to seek consensus on the latest developments. Additionally, the nations will intensify their efforts on military crisis communications. The US had initially proposed extending their trade truce with China, which, according to US Treasury Secretary Scott Bessent, will remain in effect until January.

The White House also addressed concerns over rare earth supply, with China agreeing to import at least 10 million metric tons of coal from the US in 2027 and 2028. While discussions on Iran's nuclear ambitions and the Iran war were mentioned during the talks, no major breakthroughs were reported. President Trump is expected to attend the Asia-Pacific Economic Cooperation summit in November and the G20 summit in December, both hosted in the US.

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

This story

This is one outlet's version. Read the fullest account.

Read the original at dw.com →

More in AI

How to Stop an AI Agent Infinite Loop

A simple planning request triggered 105 calls to the same goal tool. The system did not crash immediately. It kept spending tokens until the turn limit stopped it, without producing a useful answer.

  • Goal tools improperly exposed globally, causing infinite loop
  • Implemented explicit owning nodes, turn limits, call limits
  • Thorough regression testing prevented similar issues

What an Over-Engineered Parity Classifier Taught Me About Representation

I Revisited My Parity Paper — and Found the Representation Was the Real Story A while ago, I built a deliberately over-engineered classifier for one of the easiest problems in computer science: Is an…

  • Over-engineered parity classifier achieved 84.26% test accuracy with stricter evaluation protocol
  • Masking least significant bit reduced validation accuracy to random guessing (48.15%)
  • Representation-dependent structure in wavelet transform crucial for parity recovery

Self-Replicating Prompt Injections Turn Agent Context into an Open Relay

Most developers still treat prompt injection as a leakage problem. Someone types an adversarial string into your support bot, confuses the instruction hierarchy, and tricks the model into leaking an…

  • Self-replicating prompt injections possible in autonomous agents
  • GPT-Red framework used to demonstrate vulnerability
  • Four mitigation strategies recommended to prevent spread

Sentinel: The Automated Adversarial Testing Harness for LLM Applications

This is a submission for the MLH x DEV Writing Challenge What I Built AI products are shipping faster than anyone is testing them.

  • Sentinel is automated adversarial testing tool for LLM applications
  • Evaluates responses with bias-aware LLM judge, produces scored report
  • Demonstrated on vulnerable ShopCo chatbot exposing System Prompt Leakage

Microsoft Took the ‘Copilot+ PC’ Brand Out Behind the Shed

Zac Bowden, Windows Central: In 2024, Microsoft and Qualcomm kicked off a new category of Windows hardware it called “Copilot+ PCs” that set the baseline for a new wave of AI-first computers, and…

  • Microsoft discontinued "Copilot+ PC" branding by 2026.
  • Surface PCs from 2025 retained the Copilot+ label.
  • Focus shifted to producing excellent AI devices over branding.

More from Saturday 26 September →