A new phishing campaign is using WhatsApp messages to deliver malware, according to researchers at Microsoft. The attackers are attempting to trick users into installing malicious Visual Basic Script (VBS) files. “The campaign relies on a combination of social engineering and living-off-the-land techniques,” Microsoft says.

Over the past month, the cybersecurity community has published isolated reports detailing disparate attacks by the North Korean state-aligned threat group Shifty Corsair (also known as FAMOUS CHOLLIMA). While individual vendors have documented specific supply chain poisons or targeted spear-phishing campaigns, the Threat Fusion Cell (TFCTI) at BlueVoyant has synthesized these findings to reveal a much larger, coordinated offensive.

From AI agents leaking internal data to coordinated global malware campaigns — here is everything that happened in AI cybersecurity between April 7 and April 21, 2026, with detailed attack paths for each incident. The fifteen days following April 7, 2026 produced six distinct AI-related security incidents spanning internal data exposure, supply chain exploitation, autonomous malware generation, coordinated multi-vector attacks, model leak fallout, and documented AI agent control failures.

Why the next Log4Shell will be won or lost in the first 72 hours—and what a modern zero‑day workflow looks like. Every security team remembers where they were when Log4Shell dropped. A quiet Friday afternoon in December 2021 turned into a weekend of war rooms, emergency patches, and executive updates. Years on, the Log4j fallout still shows up in breach reports—a stubborn reminder that zero‑days don’t end when the news cycle does.

Most application migration projects fail the same way: someone picks a single strategy for the entire portfolio, then tries to force every workload into it. Lift-and-shift everything to meet a data centre exit deadline. Refactor everything because someone read a cloud-native manifesto. Retire nothing because no one wants to make the decision. AWS’s 7 Rs framework exists to prevent that.