ainewsblitz.com

Breaking

Researcher Says Local Uncensored Qwen Model Helped LSASS Dumper Evade Two EDR Products

  • Security
  • AI Agents
  • Foundation Models

A Project Black researcher says a locally run, uncensored Qwen3.8-27B derivative modified a Windows LSASS credential-dumping utility until it produced no detections across two lab EDR products. The September 24 test points to a practical risk for defenders: local open-weight models with stripped-down refusals can speed up iteration on offensive tooling, although the tested EDR vendors were not named and the result has not been independently reproduced.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.

$20
Read this article
$29/month
Unlimited — all 27,165 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year