Adversarial AI is on the rise: What you need to know

The Google Threat Intelligence Group and industry sources report a new phase of adversarial AI in which malware families are using large language models at runtime to generate, obfuscate, and adapt malicious code, enabling dynamic evasion and more autonomous attacks; the trend is supported by a maturing underground marketplace of AI-powered offensive tools that lowers the barrier to sophisticated phishing, data-exfiltration, and credential-theft campaigns and complicates detection for traditional signature-based defenses.