Microsoft AI Security, AI-Cyber-1-Flash And Project Perception

Should We Worry They’re Being Advertised As Less Expensive Than The Competition?

Microsoft have done a yeoman’s job with Windows Defender, going from a deservedly ridiculed attempt at malware protection to putting a serious dent in third party competition because of how effective it has become.  They are now delving into the LLM enhanced security market, with AI-Cyber-1-Flash and Project Perception. 

AI-Cyber-1-Flash is not so much active malware protection as it is a software vulnerability analysis tool, and if anyone knows software vulnerability it’s Microsoft.  From the sounds of the article at Ars Technica, the model is training on Microsoft’s decades of patching and analysis as well as processing the roughly 1 trillion security signals triggered by their 1.6 million customers every day.  This is part of what allowing telemetry to phone home on Windows 11 is about.

The other announcement is a collection of AI-powered agents called Project Perception and is intended to automate at least some of the five steps of the NIST cybersecurity framework.  Microsoft expects Project Perception to automate a fair amount of “finding vulnerabilities, investigating them to determine their risk, and taking corrective actions“.  

In less LLM filled news, AIDA has a new update.  They didn’t provide any details AMD’s upcoming Ryzen 6 family of Threadrippers, but they do offer at least some support for them.