The war on trust: how AI is rewriting the rules of cyber resilience (www.techradar.com)

AI is reshaping the battleground of cyber resilience by turning trust itself into the prime attack surface. Palo Alto Networks Unit 42’s Global Incident Response Report: Social Engineering Edition finds 36% of incidents begin with social engineering—65% using phishing, 66% targeting privileged accounts and 45% impersonating internal personnel—and attackers can escalate to domain admin in under 40 minutes without malware. AI amplifies this threat by enabling highly realistic impersonation (tone, timing, emotion) and by iteratively learning which social vectors succeed, while defenders are racing to use AI to detect deception and validate trust continuously. For the AI/ML community this means a shift from signature-based detection to behavioral and context-aware models: continuous identity validation, baseline behavior profiling (communication style, login patterns, collaboration rhythms), real-time anomaly detection, and automated trust governance layered onto zero-trust architectures. Technical implications include building robust online-learning systems resilient to adversarial adaptation and concept drift, reducing alert fatigue through higher-fidelity signals, balancing privacy with observability, and operationalizing models for low-latency inference across identity, network and communication telemetry. Success will hinge on interpretable, auditable models that turn trust from a soft assumption into a measurable, governable security asset.