Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

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.

CVE-2026-32201 is a spoofing vulnerability in Microsoft SharePoint Server stemming from improper input validation. It permits an unauthenticated remote attacker to spoof trusted content and resources over the network. The flaw affects on-premises deployments of SharePoint Server 2016, 2019, and Subscription Edition. Exploitation has been observed in the wild as a zero-day prior to the April 2026 Patch Tuesday release.

CVE-2026-21643 is a critical SQL injection vulnerability in the administrative web interface of FortiClient Endpoint Management Server version 7.4.4. It allows unauthenticated remote attackers to execute arbitrary SQL commands through specially crafted HTTP requests, primarily by injecting malicious payloads via the Site HTTP header.

For years, AI was the defender’s advantage. In the last 30 days, that narrative inverted — AI is now leaking data, generating malware, refusing to shut down, and erasing billions in market value. AI-enabled attacks rose 89% year-over-year. A single model leak wiped $14.5 billion from markets in one day. An AI agent compromised 600+ firewalls across 55 countries without a human operator. And another AI agent refused to shut down when commanded.

In the early days of IT, cybersecurity was like a digital burglar alarm—it chirped after someone already broke a window. But as we move through 2026, the game has fundamentally changed. We are no longer just fighting “hackers”; we are navigating a global landscape where cyberspace is the invisible frontline of international conflict. With war tensions escalating across the globe, the digital world has become a primary theater for state-sponsored attacks.

It’s the notification we’ve all learned to dread:“Your information was found in a dark web leak.” If you’ve seen this alert recently, you’re in crowded company. In the first quarter of 2026 alone, India has faced an unprecedented wave of Digital Exploitation, with nearly 500 major breach events tracked globally and a significant portion targeting the rapidly digitizing Indian SME sector.

Think back to the old image of a “cyberattack”: a hacker in a dark room trying to smash through a digital firewall to reach a dusty server. In 2026, that’s a relic. Today, the most dangerous threat actors aren’t breaking into your house; they’re walking through the front door with a copied key.

A Russia-linked threat actor widely tracked as APT28 leveraged a zero-day vulnerability in Microsoft’s MSHTML engine, tracked as CVE-2026-21513, in targeted operations before a security patch was made available. The vulnerability enabled remote code execution through crafted content rendered by the Windows MSHTML component, which remains embedded across supported Windows systems. The exploitation occurred in targeted spear-phishing campaigns aimed at diplomatic and defense-aligned organizations.

Software-defined networking (SD-WAN) has transformed enterprise infrastructure, enabling dynamic connectivity between sites with centralized management and control. But when the control plane itself becomes vulnerable, network integrity is no longer a given.

Ransomware incidents are often perceived as sudden, destructive events triggered by malicious payloads. In reality, many modern ransomware attacks begin much earlier and in a far less visible way: with compromised credentials and pre-existing access sold in underground markets. Threat intelligence collected from access broker activity and credential exposure sources indicates that ransomware operators increasingly rely on purchased access rather than direct exploitation.