New Threat: Cybercriminals Begin Mass Deployment of AI Agents to Bypass Traditional Firewalls

Cybersecurity researchers have detected a sharp spike in attacks controlled not by humans, but by autonomous malicious AI programs. Legacy static defense systems – even those utilizing signature-based analysis – are increasingly failing to detect these threats.

How does it work?

Once inside a network, the malware doesn’t immediately start downloading viruses. Instead, it “sits quietly,” analyzing user behavior, adapting to the system administrator’s working hours, and masking its traffic as routine requests to legitimate cloud services (such as Google Drive or Notion).

Recommendations from leading vendors (Palo Alto, Check Point, Fortinet):

• Complete abandonment of signatures – defense must rely solely on behavioral analysis (Behavioral AI). The system should react to anomalous computer behavior rather than a “known virus.”
• Implementing the XDR – collecting logs simultaneously from endpoints, networks, and clouds into a single analytical dashboard.
• Regular Security Awareness training – since AI now generates flawless phishing emails without grammatical errors and tailors them to the specific company context, the human factor remains the primary point of entry.