Generative AI

2026-07-0111 minMandiantmedium

Google Threat Intelligence Group analyzes the evolution of the pro-Russia influence ecosystem four years into the full-scale invasion of Ukraine, identifying a pivot from war-focused operations back to global strategic objectives targeting the West, NATO, and the EU. The ecosystem comprises six interconnected components — overt media, covert IO campaigns, hacktivism, cyber espionage, government direction, and outsourced proxies — that cross-promote and amplify narratives. Key trends include the increasing use of generative AI for content creation, the blending of cyber espionage with influence operations via hack-and-leak tactics, and the outsourcing of capability development to contractors like NTC Vulkan for plausible deniability.

2026-05-136 minMandianthigh

Google Threat Intelligence Group (GTIG) reports an escalation in adversaries leveraging generative AI for vulnerability discovery, autonomous malware orchestration, and defense evasion. Notable developments include the AI-assisted discovery of a zero-day 2FA bypass, the PROMPTSPY Android backdoor utilizing the Gemini API for autonomous UI navigation, and supply chain attacks by TeamPCP targeting AI dependencies like LiteLLM to extract cloud secrets.

2026-05-134 minCisco Talosinfo

Generative AI enables defenders to rapidly deploy highly adaptive honeypots that simulate complex environments like Linux shells or IoT devices. By leveraging LLMs to generate plausible responses to attacker inputs, organizations can deceive automated AI-driven attacks, shifting the defensive strategy from passive detection to active manipulation and intelligence gathering.

2026-05-134 minHuntressmedium

Threat actors are increasingly utilizing real-time deepfake technology to conduct sophisticated identity-based social engineering attacks over video calls. While physical tests like the 'three-finger test' can currently expose cheaper AI overlays due to object occlusion rendering flaws, rapid advancements in generative AI are rendering these visual tells obsolete. Organizations must shift from relying on human detection to implementing resilient, out-of-band verification processes for sensitive transactions.

2026-05-134 minPalo Alto Networksmedium

Unit 42 researchers demonstrated that both open and closed-source LLMs remain vulnerable to prompt jailbreaking at scale using a genetic algorithm-based fuzzing technique. By systematically generating meaning-preserving variants of disallowed requests, researchers successfully bypassed content filters and model guardrails, highlighting the fragility of current AI safety mechanisms under automated adversarial variation.