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.
In-Article-1
Desktop Banner (This ad will be displayed on screens larger than 992px)
Mobile Banner (This ad will be displayed on screens smaller than 992px)
In-Article-2 Desktop Banner (This ad will be displayed on screens larger than 992p
Mobile Banner (This ad will be displayed on screens smaller than 992px)
What's Your Reaction?
Like
0
Dislike
0
Love
0
Funny
0
Angry
0
Sad
0
Wow
0
Carver
Carver who carves the news
