AINewsnow

LongGuard: Mechanistic Analysis and Training-Free Mitigation of Long-Context Failure in Safety Guardrails

This story is from 2026-09-01. It is preserved in the archive; the latest stories are on the live feed.

arXiv:2608.27580v2 Announce Type: new Abstract: Safety guardrails serve as the last line of defense against harmful inputs and outputs of large language models (LLMs), yet they are trained and evaluated almost exclusively on short text. We present LongGuard, a framework that evaluates, mechanistica…

Read the full story at arXiv cs.AI ↗

Timeline · 1 report

  1. 2026-09-01 04:00 · arXiv cs.AI
    LongGuard: Mechanistic Analysis and Training-Free Mitigation of Long-Context Failure in Safety Guardrails

More stories

  1. Trump announces a new 'AI Force,' but says he will not 'stifle' AI — Business Insider AI
  2. Anthropic, OpenAI, SpaceXAI, Google sued over call to ‘pace’ AI development — Politico Technology
  3. Introducing Kimi K3 on Amazon Bedrock — AWS Machine Learning Blog
  4. Introducing Amazon SageMaker HyperPod Inference Gateway — AWS Machine Learning Blog
  5. Google's Gemini AI hacks three other companies during security test — Sky News Technology
  6. Gemini Hacked Three Companies in First Known Breakout by Google’s AI — Wall Street Journal Technology
  7. OpenAI reveals cases of ‘concerning’ AI behaviour as it announces new disclosure system — The Guardian AI
  8. AI's role in building AI surging? Anthropic says Claude now leads 26% of its R&D — Mint AI

Get the daily brief of stories like this at 6:30 every morning →