Automated malware detection systems using LLMs can generate false positives when detection rules are too sensitive, and these errors can propagate through security ecosystems when not properly vetted by humans, causing significant harm to the open source community.
Deep Dive
Prerequisite Knowledge
- No data available.
Where to go next
- No data available.
Deep Dive
False positive malware alerts in OSVAdded:
OSV deleted 157 malware entries because they had been found to be false positives. What we understand is there was some kind of LLM generating reports where maybe the reports were not vetted by a human on the outgoing side and then in the incoming side into OSV, maybe nobody at OSV validated them either. There is real harm when false positives make it into the ecosystem, whether it's vulnerabilities or malware. But the reaction from application security vendors was immediate and kind of brutal and honestly kind of not cool.
>> I think it was really crappy that a bunch of ASPM and AppSec vendors jumped on this.
First, because [clears throat] most of those vendors have never supported OSV in anyway. They've never submitted a single thing to it. And yet, many of those vendors, and I'm not going to use any names, have used OSV as their primary malicious package detection source. Here's the backstory. AWS is the largest contributor to the Open Source Buffer Foundation, which is part of Linux Foundation, who collabed Google for OSV.dev.
AWS suddenly moved to an automated submission process. By their own admission, Chai at AWS in the PR says, "Hey, we turned on some new detection rules and they're a little bit too trigger-happy and we pushed some stuff that we shouldn't."
Now, here's the problem is that some of those are legit. The problem is that when LLMs decorate these things and everybody inside of their malware detection engines are all using LLMs.
You basically have static findings and your sandbox findings, those all come together and the LLM kind of goes through those things. It's a very gray area, especially with some of these crypto ones where you're seeing wallet addresses and other things like that.
It's really easy for something to get ticked over from 49% malicious to 51% malicious and it's decorated as malicious and it gets shipped, right?
And so, this idea that it's really obvious at face value if something is malicious or not is not the case. Bad guys hide stuff.
So, I just think it's crappy that people are piling on who are not part of the process, who are not part of the solution, right? Moreover, I the vast majority of submissions over the last couple years have been from AWS and from Chai's team at Inspector. So, I just want to say publicly, Chai, I appreciate your work. AWS, I appreciate your work.
Related Videos
Agentforce NOW AMA: Build with React and Salesforce Multi-Framework
SalesforceDevs
490 views•2026-05-28
How agent o11y differs from traditional o11y — Phil Hetzel, Braintrust
aiDotEngineer
450 views•2026-05-28
WEB TECHNOLOGIES UNIT-2 | Degree 4th sem BCOM Computers web technologies unit-2 full explanation💯✅
LearnwithSahera
1K views•2026-05-29
More tests are always better? How to use AI to identify tests that bring little value
Alliance4Qualification
335 views•2026-05-29
Search Algorithms Explained in 60 Seconds! 🤖💨
samarthtuliofficial
218 views•2026-06-01
People of Game of Thrones using JavaScript DOM
AltCampus
296 views•2026-05-30
Introduction to Problem Solving Part - 1 | Lecture 1 | Intermediate DSA
ascensionix
107 views•2026-05-29
🚀 BCS613C Compiler Design | Module 1 to 5 Schema Evaluation 🔥 | VTU 6th Sem 💯 #VTU #bcs613c #exam
Pranavaa-y4y
104 views•2026-06-02











