Important read, thanks for sharing. Do you think AI-driven phishing will soon make traditional email filters obsolete, or can cybersecurity tools keep up with this level of personalization?
Based on my conversations at Black Hat 2025, this article captures a critical trend that security leaders are grappling with right now. The shift from "spray and pray" to hyper-targeted attacks is real and accelerating.
What stood out from my discussions with experts like Jim Dolce from Lookout is that AI isn't just making phishing more personalized—it's expanding the attack surface beyond email. His team demonstrated how they created a convincing voice phishing attack in just 15 minutes using AI. The synthetic voice was so realistic that even his wife couldn't distinguish it from his real voice.
The Famous Chollima example you mentioned aligns with intelligence I heard from Cristian Rodriguez at CrowdStrike. Their research shows this North Korean group has infiltrated over 320 companies—a 220% increase—using AI-generated résumés and deepfake technology in video interviews. They've become what Rodriguez calls "the most GenAI-proficient adversary."
But here's what the article doesn't capture: defenders actually have some advantages in this AI arms race. Ryan Fetterman from Splunk showed how they're using AI for "model-in-the-loop threat hunting," achieving 80% accuracy in detecting malicious PowerShell scripts while reducing analysis time from five minutes to two seconds.
The real challenge isn't just recognizing sophisticated phishing—it's that organizations can no longer train their way around these attacks. As Dolce pointed out, "You cannot train your way around an AI-generated exploit. The AI-generated exploit is way too smart."
The defense now requires AI-powered solutions that can analyze communication patterns in real-time. Lookout's approach achieves 98% accuracy by asking AI models to evaluate whether messages are legitimate, matching the sophistication of AI-powered attacks with AI-powered defenses.
The question isn't whether your defenses are ready—it's whether you're fighting AI with AI.