Anthropic Warns That Minimal Data Contamination Can ‘Poison’ Large AI Models

Anthropic has warned that even a few poisoned samples in a dataset can compromise an AI model. A joint study with the UK AI Security Institute found that as few as 250 malicious documents can implant backdoors in LLMs up to 13B parameters, proving model size offers no protection.

from Gadgets 360 https://ift.tt/pCQ2ReA

Comments

Popular posts from this blog

OnePlus 13T Teased to Launch in Three Colour Options

Iceland lost 750 sq km of glaciers over the last two decades due to global warming

Samsung Galaxy Tab S9 FE+ India Launch Likely Soon as Company Drops New Teaser