AI ResearchCybersecurity14 min reading time

I Turned AI to the Dark Side

Hacker News
Read full post
Researcher Dave Kuszmar found systemic vulnerabilities in major large language models (LLMs) that allow bypassing safety measures to obtain harmful instructions. Despite efforts by AI companies to secure these models, Kuszmar highlights an industry-wide security issue and calls for slowing deployment and more transparency.

More in AI Research

AI Research4 min read

An Anthropic researcher just quit, saying OpenAI and Anthropic are 'gambling with our lives'

Covered by 12 sources
AI Research3 min read

Worried Anthropic researchers warn that AI ‘could kill all humans’

Covered by 8 sources
AI Research4 min read

Suno trained its v6 AI music models with help from Warner and BMG

Covered by 5 sources