Alerts
Events
DCR
Explore Cyware Products
Alerts
Events
DCR
Go to listing page
New Attack Reveals the Inner Workings of OpenAI's and Google's Closed Models
Malware and Vulnerabilities
March 13, 2024
The Register
This attack has been able to extract the entire projection matrix of OpenAI's language models for a cost ranging from a few dollars to several thousand, depending on the model's size and the number of queries.
Read More
OpenAI
Google Compute Engine
Large Language Models (LLMs)
Projection Matrix
Attack Technique
Publisher
Previous
Eye Security Raises $39 Million in Series B Funding
Companies to Watch
Next
DARPA Picks Seven Small Businesses for AI Cyber Challen ...
Innovation and Research