In 2025, the ransomware ecosystem did not just grow — it exploded. Driven by the mass adoption of artificial intelligence among cybercriminal groups, the number of victims posted to dark web data leak sites rose 58% year-over-year, reaching 7,515 claimed victims by the end of December. December alone saw 814 attacks — a 42% year-on-year increase. We are witnessing a fundamental shift: AI has transformed ransomware from a specialist criminal enterprise into a scalable, semi-automated industry accessible to anyone with intent and an internet connection.
The Numbers Don’t Lie: Ransomware’s Record-Breaking Year
Multiple threat intelligence firms independently confirmed 2025 as the most destructive ransomware year on record. GuidePoint Security’s GRIT team tracked 2,287 unique victims in Q4 2025 alone — the highest quarterly count ever recorded. Recorded Future logged 7,200 publicly reported attacks, up from 4,900 in 2024, a 47% increase. Comparitech’s full-year tally stood at 7,419 attacks globally, a 32% rise over the prior year.
Check Point Research documented a 60% increase in ransomware attacks between December 2024 and December 2025. IBM’s 2026 X-Force Threat Intelligence Index reported a 49% surge in active ransomware groups year-over-year, driven partly by the collapse of barriers to entry — a collapse fuelled, in large part, by AI. Manufacturing led all sectors as the most targeted industry, accounting for nearly 29% of all attacks, followed by technology, retail, and healthcare.
AI as the Great Equaliser for Cybercriminals
For decades, the technical ceiling of ransomware was a natural barrier. Developing functional encryption payloads, evading endpoint detection, and coordinating exfiltration required genuine programming skill. That ceiling is now gone.
AI tools — including large language models (LLMs), code-generation assistants, and agentic frameworks — now allow even low-skill threat actors to rapidly prototype, iterate, and deploy malware. According to IBM X-Force, AI has enabled attackers to identify weaknesses in public-facing applications 44% faster, driving a corresponding surge in vulnerability-exploitation as the leading initial access vector (accounting for 40% of X-Force-observed incidents in 2025).
Malwarebytes’ 2026 threat report described 2025 as the year cybercrime began its shift toward an AI-driven future, noting the first confirmed cases of AI-orchestrated attacks alongside deepfake-enabled social engineering and AI agents that outperformed human researchers at discovering vulnerabilities. A 2025 MIT study found that an AI model using the Model Context Protocol (MCP) achieved domain dominance on a corporate network in under an hour with no human intervention, evading endpoint detection and response (EDR) tools via on-the-fly tactic adaptation.
“The 2025 State of Ransomware Survey reinforces that legacy defences can’t match the speed or sophistication of AI-driven attacks. Time is the currency of modern cyber defence — and in today’s AI-driven threat landscape, every second counts.”
— CrowdStrike, 2025 State of Ransomware Report
CrowdStrike’s survey of global organisations found that 76% struggle to match the speed and sophistication of AI-powered attacks, while 85% say traditional detection methods are becoming obsolete. Nearly half — 48% — cited AI-automated attack chains as the single greatest ransomware threat they face today.
Case Study: FunkSec — The First Widely Confirmed AI-Built Ransomware
No case illustrates the AI-ransomware convergence more starkly than FunkSec. The group emerged publicly in late 2024 and, within a single month — December 2024 — claimed more victims than any other ransomware group globally, publishing over 85 targets on its dark web data leak site. By the close of its active campaign (March 2025), it had compromised at least 113 confirmed organisations across the US, India, Italy, Brazil, Israel, Spain, and Mongolia.
What made FunkSec remarkable was not its victim count — it was how it was built. Check Point Research’s analysis concluded that the group’s encryptor and attack tooling were likely developed with AI assistance, enabling rapid iteration despite what analysts described as the author’s apparent lack of technical expertise. FunkSec’s lead developer explicitly told Check Point researchers they were a developer and not a coder — relying on generative AI and collaborators to translate intent into functional malware.
“More and more, we see cybercriminals leveraging AI to develop malicious tools. Generative AI lowers barriers and accelerates malware creation, enabling cybercriminals to adapt their tactics faster. By reducing the entry threshold, AI allows even less experienced attackers to quickly develop sophisticated malware at scale.”
— Marc Rivero, Lead Security Researcher, Kaspersky GReAT
Technical Breakdown: How FunkSec Works
FunkSec V1.5 is written in Rust — a language chosen for its performance characteristics and resistance to static analysis. Upon execution, the malware follows a structured attack sequence:
- Privilege Escalation: The binary checks for and attempts to obtain administrative/root privileges before proceeding.
- Defence Suppression: It terminates over 50 hard-coded processes and services — including Windows Defender, browsers, media players, and backup agents — to clear the path for encryption and evade detection.
- Shadow Copy Deletion: Volume Shadow Copies (VSS) are deleted, destroying Windows’ built-in recovery mechanism.
- Encryption: Files across all local drives are recursively encrypted using the ChaCha20 cipher combined with Poly1305 MAC (via the orion-rs cryptographic library v0.17.7), operating on 128-byte blocks. Encrypted files grow approximately 37% in size and are appended with the .funksec extension.
- Ransom Note Deployment: A randomised ransom note (README-[random].md) is dropped in every affected directory.
- Data Exfiltration: A password-gated mode activates aggressive data exfiltration, with stolen data later sold on dark web markets for $1,000–$5,000.
Kaspersky’s GReAT team noted that FunkSec’s ancillary tools — including a DDoS module and password generator — showed clear signs of LLM-assisted code synthesis. Even the code comments were written in flawless English, contrasting sharply with the basic English the group used on its own platforms — a tell-tale sign of AI involvement flagged by multiple independent researchers.
FunkSec operated a Ransomware-as-a-Service (RaaS) model, charging affiliates just $10,000 per attack — dramatically undercutting the market while maximising volume. Avast later released a free decryptor for victims following the group’s dismantlement in mid-2025, but the blueprint FunkSec established has since been replicated across the broader ransomware ecosystem.
Real-World Impact: Major Attacks of 2025–2026
Qilin: The Year’s Most Prolific Ransomware Group
The Russia-linked Qilin group (also known as Agenda) dominated the 2025 threat landscape, claiming 1,034 attacks across the year — 14% of all recorded global incidents — and surpassing LockBit’s previous peak in absolute terms. By October 2025 it had already hit its 700th victim, eclipsing the prior year’s leading strain RansomHub, which totalled 547 victims across all of 2024.
Qilin employs a mature double-extortion model, stealing sensitive data before encrypting systems using AES-256 CTR or ChaCha20 algorithms. The group gains initial access by exploiting vulnerabilities in public-facing applications — notably Fortinet VPNs and Veeam Backup and Replication — before moving laterally with tools like PsExec and terminating critical backup processes. Its Rust-based variant is notably harder to analyse and detect than legacy Go-based predecessors.
Perhaps its most notorious operation was the 2024 attack on Synnovis, a UK pathology services provider serving Guy’s and St Thomas’ and King’s College NHS Trusts in London. Qilin demanded a $50 million ransom. Synnovis refused to pay. The consequences were severe: an estimated 900,000 NHS patient records were exposed, thousands of appointments and operations were cancelled, and in June 2025 King’s College Hospital NHS Trust confirmed the disruption had contributed to the death of a patient — one of the first times a ransomware attack has been directly linked to a fatality. Total costs exceeded £33 million (~$44 million).
Clop’s Zero-Day Campaign Against the NHS
In November 2025, NHS England found itself listed on the dark web leak site of ransomware group Clop, following an exploitation campaign targeting zero-day vulnerabilities in Oracle’s E-Business Suite (CVE-2025-53072 and CVE-2025-62481). Clop — operating through a data-theft-and-extortion model without necessarily deploying encryption — simultaneously claimed The Washington Post as a victim. NHS Barts Health, one of England’s largest trusts running five London hospitals, confirmed in December 2025 that Clop had accessed invoice data, prompting it to seek an emergency High Court injunction to prevent data publication.
Conduent: 15.9 Million Records Compromised
The single largest data breach linked to ransomware in 2025 involved US technology services giant Conduent, where a ransomware attack resulted in up to 15.9 million records being affected — a figure that underscores the catastrophic scale achievable when supply-chain-targeting ransomware campaigns compromise a single provider serving thousands of downstream clients.
The Technical Mechanics of AI-Enhanced Ransomware
1. Reconnaissance and Vulnerability Discovery
AI models can scan internet-exposed infrastructure, parse CVE databases, correlate exposed service versions with known exploits, and prioritise targets — all at machine speed. IBM X-Force observed a 44% increase in attacks beginning with exploitation of public-facing applications in 2025, attributed in part to AI-accelerated vulnerability discovery. Trend Micro’s 2026 security predictions warn that attackers will use AI to discover and weaponise vulnerabilities faster than defenders can respond.
2. AI-Powered Phishing and Social Engineering
Phishing remains the most common initial access vector. AI supercharges it through hyper-personalised lures, flawless multilingual copy, deepfake audio and video for CEO fraud, and automated spear-phishing at scale. CrowdStrike found that 87% of organisations believe AI makes phishing lures more convincing. Ransomware groups increasingly use LLMs to overcome language barriers, allowing non-English-speaking operators to target English-speaking organisations with native-quality social engineering content.
3. Malware Development and Rapid Iteration
As FunkSec demonstrated, AI can translate intent into functional malware code without the author possessing deep programming expertise. Beyond raw code generation, LLMs assist with obfuscation techniques, generating polymorphic variants that evade signature-based detection, and writing documentation that helps affiliates deploy correctly. IBM found infostealer malware exposed over 300,000 ChatGPT credentials in 2025 alone, signalling that AI platforms themselves have become high-value targets within the attack ecosystem.
4. Autonomous Attack Execution via AI Agents
The most alarming frontier is fully autonomous attack execution. Malwarebytes predicts that in 2026, AI capabilities will mature into fully autonomous ransomware pipelines allowing individual operators to attack multiple targets simultaneously at unprecedented scale. Trend Micro concurs: agentic AI will soon handle reconnaissance, vulnerability scanning, lateral movement, and even ransom negotiation — all without human oversight.
“The proliferation of AI-powered RaaS platforms will further democratise cybercrime, allowing even novice criminals to deploy highly effective attacks.”
— Janus Agcaoili, Senior Threat Research Engineer, Trend Micro
The Defender’s Dilemma: Confidence Is Dangerously Low
The confidence gap between attackers and defenders has never been wider. ISACA’s 2025 survey of 632 European cybersecurity professionals found that only 7% feel extremely confident their organisation could successfully navigate a ransomware attack in 2026, and just 14% feel very prepared to manage generative AI risks. CrowdStrike’s survey revealed a critical disconnect: 76% of organisations report a gap between leadership’s perceived ransomware readiness and actual preparedness on the ground.
Perhaps most troubling: 83% of organisations that paid a ransom were attacked again, and 93% had their data stolen regardless. Paying ransoms is not a strategy — it is a funding mechanism for repeat victimisation.
What Defenders Must Do: A Strategic Response
- Adopt AI-powered security tooling. Legacy signature-based detection is increasingly ineffective against polymorphic, AI-generated malware. Behavioural detection, anomaly-based EDR, and AI-driven threat hunting are now essential. CrowdStrike found 89% of organisations view AI-powered protection as critical.
- Prioritise patch velocity. Vulnerability exploitation accounted for 40% of initial access vectors in 2025. Automated patch prioritisation pipelines are now a competitive necessity.
- Harden identity systems. Enforce phishing-resistant MFA (FIDO2/passkeys), deploy privileged access management (PAM), and implement continuous conditional access policies.
- Implement immutable, offline backups. VSS deletion is standard in virtually every modern ransomware payload. Immutable, air-gapped or cloud-isolated backups are your last line of defence.
- Shrink the attack surface. Audit and eliminate unnecessarily exposed services. Every public-facing application is a potential AI-enumerated target.
- Adopt continuous monitoring. AI agents can achieve domain dominance in under an hour. Detection and response dwell time must compress dramatically. Network segmentation and real-time lateral movement monitoring are non-negotiable.
- Simulate real-world attacks. Regularly test defences against the specific TTPs of active groups including Qilin, Akira, and Clop using threat emulation platforms.
The Road Ahead: 2026 and Beyond
The trajectory is unambiguous. IBM X-Force expects adversaries to automate complex tasks including reconnaissance and ransomware deployment as multimodal AI models mature, driving faster-moving and more adaptive threats. Recorded Future predicts 2026 will mark the first year new ransomware actors outside Russia outnumber those emerging within it — reflecting the truly global democratisation of AI-enabled cybercrime.
Trend Micro frames it clearly: ransomware is evolving from a disruptive event into a systemic issue — one where agentic AI handles the full attack chain from target selection to ransom negotiation with minimal human intervention.
The good news? AI is equally available to defenders. The organisations that will weather this wave are those that match the attacker’s automation with their own: AI-driven detection, automated response playbooks, and a security culture that treats the threat with the urgency it deserves.
Key Takeaways
- Ransomware attacks rose 58% in 2025 to the highest level ever recorded, with AI as the primary accelerant.
- FunkSec proved that AI enables even low-skill actors to build and deploy functional ransomware at scale.
- Qilin claimed 1,034 attacks in 2025; its Synnovis assault cost £33 million and contributed to a patient death.
- AI is now weaponised across the entire attack chain: reconnaissance, phishing, malware development, and autonomous execution.
- Only 7% of cybersecurity professionals feel confident their organisation could withstand a 2026 ransomware attack.
- AI-powered defences, rapid patching, identity hardening, and immutable backups are the minimum viable security posture.
Stay ahead of the threat. Follow V3ndta for ongoing analysis of the cybersecurity landscape, ethical hacking insights, and emerging threat intelligence.
Leave a Reply