Microsoft’s new AI agent can spot malware without human help
Readers help support Windows Report. We may get a commission if you buy through our links.
Read our disclosure page to find out how can you help Windows Report sustain the editorial team. Read more
Microsoft has announced a new AI-powered tool called Project Ire, and it’s built to spot malware on its own. While it’s still early days, the results already look promising.
In a real-world test, Microsoft used almost 4,000 suspicious files, and guess what, Ire correctly identified nearly 9 out of 10 malicious files. Well, it didn’t spot everything, though. According to Microsoft’s blog, the system flagged only about 25% of the total malicious files present.
That said, the files it did catch were mostly accurate, with very few false positives. The company calls this type of classification the “gold standard” in malware detection.
Typically, it takes skilled analysts hours to reverse-engineer questionable software. They have to dig into each file, figure out its intent, and decide if it poses a threat. But that process doesn’t scale, especially when threats number in the thousands.
Project Ire tries to solve this by breaking malware analysis into smaller steps. It reasons in layers, which helps prevent overload and boosts accuracy. The AI also pulls from a wide range of tools, including sandbox environments, decompilers, memory analysis, and even internal documentation.
Microsoft is planning to integrate it into Defender to speed up and scale threat detection. Meanwhile, Google is also moving in a similar direction. Its AI-powered bug-hunting agent, shown off earlier this year, recently spotted nearly 20 vulnerabilities.
Microsoft admits Project Ire still has a way to go, but it’s a glimpse into how defenders could get faster, smarter, and less dependent on manual work, especially when dealing with modern malware at scale.
