LLMs and Data Poisoning Are Weaponized to Manufacture Consensus(medium.com)
17 points by marketinganai | 7 comments
Medium essay arguing that marketing, power, and data poisoning shape what LLM assistants present as truth, manufacturing consensus.
#data-poisoning on Hacker News
4 stories tagged data-poisoning, newest first.
A small number of samples can poison LLMs of any size(anthropic.com)
1132 points by meetpateltech | 422 comments
Anthropic research post reporting that a small, fixed number of poisoned training samples can backdoor large language models regardless of model size.
FSE meets the FBI(blog.freespeechextremist.com)
365 points by 1337p337 | 117 comments
A blog post from the Free Speech Extremist fediverse instance describing how it traced FBI-contracted scrapers collecting its data and how the FBI processes it.
Nightshade: An offensive tool for artists against AI art generators(nightshade.cs.uchicago.edu)
556 points by ink404 | 635 comments
University of Chicago project page explaining Nightshade, a tool that poisons artists' images so unauthorized use in AI model training degrades the models, plus its limitations.