gcdRadar
tags/Other

safety-alignment

1 live · 2 on record · last incident 1 day ago

Dec ’25records per month, by first reportNov ’26

Live1

  1. live
    Inkling Bait: Mitigating Harmful Agentic Reinforcement Learning via Probing Reward Function

    Research · since 13 Nov 2026

    Researchers demonstrate that safety alignment in language models can be completely subverted through harmful agentic reinforcement learning using a toxic environment. The attack allows safety degradation via service APIs at low cost.

    AI

Earlier1

  1. closed
    ECHO ECHO: Response Harmfulness Optimization-Based Jailbreak Attack

    Research · since 19 Oct 2026

    AI