AI Dynamics

Global AI News Aggregator

About

Data-Poisoning Attacks Threaten LLMs Regardless Size

New research with the UK @AISecurityInst and the @turinginst
: We found that just a few malicious documents can produce vulnerabilities in an LLM—regardless of the size of the model or its training data. Data-poisoning attacks might be more practical than previously believed.

→ View original post on X — @anthropicai