Recent Stories
American Biotech — The Future Can’t Be Won Without It.RSI is AI Building AI. What DC needs to know about it. Keeping the Human in Medicine: Two AI CoLab Researchers on What Machines Can’t ReplaceSomeone Is Hacking the Freezers. Pay Attention.Thinking of Running in 2028? Start HereWhat the Lame Duck Session Has to Do With the Ultimate 21st-Century Power Tool: Robots

The Wave

Senatein sessionHousein session
Refreshed just now
News

Your AI Knows It's Being Tested: How METR's Investigator Agents Spot Alignment Evals With Perfect Accuracy

Why it’s here: Covers AI alignment evaluation integrity and frontier model behavior under testing, a technical development bearing on AI safety evaluation policy the DC audience tracks, though it is a secondary write-up of a METR note rather than a policy event.

Trace across the board

Item

  1. 5d agoDEV Community · research by METRYour AI Knows It's Being Tested: How METR's Investigator Agents Spot Alignment Evals With Perfect AccuracyMETR research finds that investigator agents increasingly distinguish alignment evals from capability evals, mainly by noticing environments look fake or that tasks are honeypots.
    2 articles
    1. 6d agoMETREarliestMeasuring alignment eval tells
    2. 5d agoDEV CommunityPickupYour AI Knows It's Being Tested: How METR's Investigator Agents Spot Alignment Evals With Perfect Accuracy
    AI5 linked

Coverage

2 outlets · first seen 6d ago

Connected

2 links

About

  • Claude Opus 4.7
  • Claude Opus 4.7
  • Petri
  • GPT-6 Astra

Analytics cookies help us see how the site is used. They only run if you accept. Privacy Policy