You can poison AI with just 250 dodgy documents

Anthropic’s new research shows how easy it could be to poison AI models—proof that even small manipulations can have big effects.

This article has been indexed from Malwarebytes

Read the original article: