Urgent.News

What's breaking now, across thousands of outlets.

AI

DARPA wants AI to find hidden flaws in digital markets that adversaries could exploit to wreak economic havoc

DARPA is studying how the rules governing online markets could be manipulated to cause disruption, rather than just how their systems could be hacked.

The Defense Advanced Research Projects Agency (DARPA) is seeking artificial intelligence (AI) tools to identify hidden weaknesses in digital markets that adversaries could exploit to cause economic chaos. The agency, part of the Pentagon, is developing the Intrinsic Market Resilience (IMR) program to protect critical systems like electronic markets, which have become faster, more efficient, and increasingly interconnected.

DARPA warns that while AI can make markets more vulnerable, it can also help defend against potential disruptions. The program aims to leverage AI, particularly large language models, to create computer models of market operations and interactions, identifying design flaws and proposing improvements. DARPA emphasizes the need to protect against disruptions that could cause economic damage or instability, rather than financial gain for individual actors.

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

AI Can Write Code, But What About Us?

🇮🇩 Originally written in Indonesian. This English version was AI-assisted and adapted for a more natural reading experience. It is not a literal translation. Open Link Introduction Just yesterday.

  • AI tools are essential for developers to understand, not just rely on
  • Developers must control AI usage, aligning with goals and evaluating results
  • Adaptation to AI is necessary; developers should incorporate it into workflows

Reproducible LLM Evals: Record the Run, Not Just the Score

Your eval suite scored 87% on Tuesday and 84% on Thursday. Nobody changed the prompt. Was it the model, the dataset, the grader, or noise?

  • Record the entire run, not just the score
  • Temperature does not guarantee repeatability
  • Measure and report the noise in generation

More from Sunday 8 November →