Security News

Cybersecurity news aggregator

Analysis April 21, 2026 15 min read By Security News

The ransomware economics shift — and the uncomfortable question about victims

Attacks keep climbing. Payments keep falling. Most of the story is straightforward — better backups, harder sanctions, smarter refusals. But a newer strand running through recent DOJ filings is harder to talk about: sometimes the "victim" is part of the operation.

Short answer

Volumes are up, but payment rates have collapsed. In 2022, 79% of identified ransomware victims paid. In 2025, only 28% did. Chainalysis recorded the first-ever year-over-year drop in ransom revenue in 2024 ($813M, down 35% from $1.25B in 2023).

Ransomware has also become a money-laundering infrastructure. North Korea stole $2.02B in crypto in 2025 (≈13% of its GDP). Russian exchange Garantex processed $100M+ in illicit flows before OFAC re-sanctioned it and seized its successor Grinex in August 2025.

The angle most coverage misses: victims and perpetrators are no longer cleanly separable. In November 2025, ransomware negotiators at DigitalMint and Sygnia — two of the most trusted US incident-response firms — were indicted for working both sides of attacks they were hired to resolve.

AI is part of the picture, but not yet decisive. FunkSec became the most active ransomware group of December 2024 using AI-assisted code, and jailbroken models like WormGPT 4 sell for $220 lifetime. The practical effect so far is democratisation — more amateurs shipping functional ransomware — not more sophisticated attacks. See the AI dimension section for detail, and our companion article Is AI speeding up cyber attacks? for the phishing and access-side picture.

If you read nothing else

Most Icelandic organisations will never face a corrupt ransomware negotiator or a state-sponsored laundering operation. They will face a stolen credential, an unpatched service, and a Tuesday morning. Three controls defeat most of that — and they are cheap:

  1. Phishing-resistant MFA (FIDO2 hardware keys) on every admin, email and finance account. Not SMS.
  2. Offline, immutable backups, restore-tested monthly — not annually, not "we have a backup somewhere."
  3. A written ransomware-payment policy approved by the board before an incident, not during.

The rest of this article explains why those three defences are correct and what the global picture looks like. If you already have them in place, you are doing better than most.

Dataset 1 — Our live data

Ransomware victim disclosures tracked here

We ingest ransomware leak-site postings from ransomware.live every 30 minutes. The numbers below reflect our own stored observations since February 2026.

Most active groups

Targeted sectors

Monthly disclosures

What it shows

Qilin and TheGentlemen dominate current operations. Manufacturing and Technology are the top targets — sectors with the highest cost-per-hour of downtime and the most pressure to pay quickly. Our data does not cover Iceland well because ransomware.live relies on the groups' own leak sites; Icelandic victims who negotiate privately or have data withheld never appear in these counts.

Dataset 2 — Chainalysis on-chain tracking

Total ransom paid per year (global)

Chainalysis tracks cryptocurrency flows to known ransomware wallets. Figures are revised upward over time as more incidents are attributed.

What it shows

2023 was the record year at $1.25B. 2024 saw the first-ever year-over-year drop — down 35% to $813M. 2025 is tracking around $820–900M. Much of the 2024 decline happened after July, following the LockBit disruption (H2 LockBit payments fell ≈79%), the Chipmixer and Sinbad mixer seizures, and international law-enforcement pressure on affiliate programs.

Dataset 3 — Coveware incident-response telemetry

Percentage of victims who actually pay

Coveware (acquired by Veeam) reports quarterly data from real ransomware engagements.

What it shows

The collapse from 78.9% paying (2022) to 28% (2025) is the most consequential shift of the decade. Better backups, board-level refusal, OFAC sanctions deterrence, and the growing assumption that paying doesn't prevent a leak — all of these push the rate down. When victims do pay, the average payment has climbed: $1.13M average in Q2 2025 (+104% from Q1). Attackers compensate by targeting fewer, richer victims.

The supply side

The $500 problem: Initial Access Brokers

You rarely need to "hack" anything to run ransomware in 2026. You buy the access, pre-authenticated, on a dark-web forum.

SMB RDP access
$500
base listing on Russian Market / XSS
Corporate VPN credentials
$2,871
average selling price (2024 survey)
Domain admin account
$8,167
average
RDP admin access
$9,800
average
Fortune 500 domain admin
$50,000+
high-value listing
Average bundle victim size
$2.2B
in annual revenue
The economics

For the price of a dinner, an attacker can buy an RDP session into a small Icelandic business. For the price of a used car, they can buy domain administrator access to a mid-market enterprise. The ransomware actor is no longer a skilled intruder — they are a renter of other people's intrusions. This is why segmentation, MFA on every exposed service, and continuous credential-leak monitoring matter far more than any single patch.

Supply-side commoditization, part two

The AI dimension — concrete, but not yet decisive

If 2022–2024 was the era when network access became a commodity, 2024–2025 is when ransomware code itself did. This is not the AI-hype version of the story. The measurable impact is smaller than the headlines suggest, but it is real, and it points in a specific direction.

FunkSec: the first credible AI-assisted ransomware group

In December 2024, a new group called FunkSec published 85 victims on its leak site — briefly making it the most prolific ransomware actor that month. Check Point Research identified the FunkSec encryptor as AI-assisted: written in Rust, with telltale artefacts in the build environment (paths referencing C:\Users\Abdellah\) that a seasoned malware author would have sanitised. The supposed author appears to be an Algerian amateur rather than an established crew. Ransom demands were unusually low — often $10,000 — and a significant portion of the "leaked" data turned out to be recycled hacktivism material. The group's growth was real; its technical sophistication was not.

Malicious-LLM-as-a-service

A parallel ecosystem of jailbroken models is now sold openly on underground forums. WormGPT 4 is the most visible, priced at $50/month or $220 for lifetime access. Variants include FraudGPT, WolfGPT and KawaiiGPT, largely built on top of open-weight base models (Mixtral, Grok-adjacent forks) with the safety layers removed. Operationally they produce ransomware skeletons, phishing copy, and lateral-movement snippets without guardrails. What they do not do, realistically, is outperform what a competent engineer could fine-tune themselves in an evening. The novelty is the distribution, not the capability.

The first fully documented AI-driven attack chain

In November 2025, Anthropic disclosed that a China-linked threat actor had used its Claude models to autonomously run end-to-end attacks against approximately 30 organisations — from reconnaissance through data exfiltration, with minimal human direction. The incidents were not all ransomware, but the pattern matters: mainstream, well-defended AI platforms have already been turned into attacker infrastructure at least once.

What AI has not done

The payment-rate collapse from 78.9% (2022) to 28% (2025) happened before AI could plausibly have driven it. The drivers were defender-side: better offline backups, OFAC deterrence, insurance pressure, board-level refusal policies. Attackers adopting AI have not reversed those trends, and there is no credible evidence that AI has meaningfully increased ransomware volume or average payment size. It has, so far, done one specific thing: lowered the technical threshold to ship a functional ransomware operation.

The actual shift: democratisation, not sophistication

An amateur with $500 of IAB credentials and a $220 WormGPT subscription can now assemble a working ransomware operation. FunkSec is proof of the model. The strategic implication for defenders is that the volume of low-skill attackers will increase while the quality of the malware in each attack may actually decline. Traditional signature-based detection will struggle with AI-generated variants that look different in every compile. Behavioural EDR and segmentation will still catch them — they still have to move laterally and encrypt files — but the filter-by-reputation assumption ("this is obviously amateur-hour, not a real threat") no longer holds. See our companion article Is AI speeding up cyber attacks? for the phishing and access-side picture.

The specific thing to watch for in Iceland

The defence that worked best against past ransomware in Iceland was language: most ransom notes were in broken English, most phishing was obviously foreign. AI has ended that defence. Expect ransom notes, extortion emails and negotiation chat in fluent, culturally aware Icelandic from 2026 onward. Staff training built on "the grammar gives it away" is no longer a control — it is a liability.

The laundering layer

Ransomware as money-laundering infrastructure

Ransomware is also how sanctioned states move money. This isn't a hypothesis — OFAC has documented the flows and sanctioned both groups and the exchanges that service them.

North Korea: ransomware as national revenue

Russia: Garantex and its successors

The mixer crackdown and what replaced it

Historically 10–15% of ransomware-laundering flows went through anonymising mixers. After OFAC/FBI actions against Tornado Cash, Sinbad.io and Chipmixer, that share collapsed. The 2025 replacement stack: centralised exchanges (often ones the attackers control or have captured staff at), cross-chain bridges to hop between blockchains, DeFi protocols for fragmentation, and personal wallets held for long cooldown periods before cashing out.

Why this matters for Icelandic organisations

Paying a ransom to a sanctioned group or through a sanctioned exchange is a potential OFAC violation. US sanctions attach to any USD-denominated transaction, including SWIFT transfers originating from Icelandic banks. FinCEN's revised 2021 advisory (FIN-2021-A004, re-issued 2025) explicitly warns that financial institutions facilitating ransomware payments may themselves face sanctions exposure. A compliant response requires real-time sanctions screening of the payment destination — most Icelandic firms do not have that capability in-house, and neither do most insurance brokers.

The uncomfortable part

When the "victim" is part of the operation

Read this as an industry-level risk, not a direct one. The cases below are US-documented and involve IR firms at the scale of DigitalMint and Sygnia. The direct threat to most Icelandic organisations is not a corrupt negotiator — it is an automated attacker using a purchased credential. This section matters because it shapes how insurers, auditors and regulators will increasingly treat all ransomware cases, including yours, even if no single Icelandic organisation faces an insider-run attack today.

This is the story most ransomware coverage avoids, but it matters for how boards, insurers and regulators should treat reported incidents. In 2025 the DOJ filed a series of indictments that confirm what investigators had suspected for years: the line between victim, responder and attacker is blurred, and sometimes the same person sits on all three sides.

Documented cases

A taxonomy of "victim as perpetrator"

PatternWho is the hidden actor?Evidence level
Insider leaks credentials for a cutAn employee of the victimDocumented across DOJ filings
Compromised IR / negotiation firmNegotiator working both sidesDigitalMint + Sygnia, Nov 2025
Staged attack for insurance or tax deductionVictim's own finance or IT leadershipFinCEN 2025 FTA flags the risk; public cases are scarce but investigations are ongoing
Shell "IR firm" run by the attackerThe attacker, wearing a white-hat costumeBlackCat cybersecurity-expert arrests
Paid "ransomware-as-a-service" on yourself to launder fundsVictim and attacker share the outcomeTheoretical but flagged by FATF (March 2023) and FinCEN (December 2025)
Double-charge / triple-extortion collusionSeveral actors splitting upliftFBI on Medusa; CISA StopRansomware advisory
Why this changes your incident-response plan

Single-vendor trust is no longer safe. Any ransomware response that involves a payment should include: (1) an independent secondary adviser unaffiliated with the primary negotiator, (2) OFAC sanctions screening of the destination wallet, documented before funds move, (3) chain-of-custody documentation of who controls the payment wallet at every step, and (4) a legal review explicitly covering the possibility that the attack was staged or facilitated by an insider. Auditors and insurers should treat any payment from a wallet with no prior transaction history, or to an exchange recently added to the OFAC list, as evidence requiring further investigation rather than routine loss.

The local view

Iceland in the ransomware crosshairs

Why so few Icelandic victims appear in leak-site data: most ransomware groups target by language, revenue and internet exposure — not nationality. Icelandic organisations that do get hit tend to be smaller, so any payment is below the thresholds the big leak sites bother publishing, and many attacks are resolved through quiet negotiation without public disclosure. This is about to change under NIS2: mandatory 24-hour reporting will make incidents visible whether the group publishes them or not.

What works

Defensive priorities that actually move the needle

Executive checklist

Next steps by role

For the board / C-suite

For IT / security leadership

For finance / legal / audit

Methodology & sources

Dataset 1 (our live data): ransomware.live leak-site feed, ingested every 30 minutes into our own database since February 2026. Counts exclude Icelandic victims not published by the groups themselves — which is most of them.

Dataset 2 (Chainalysis): on-chain tracking of known ransomware wallets. Figures are always revised upward over time as additional incidents are attributed; the 2025 total of ~$820M is expected to climb toward $900M once late attributions are applied.

Dataset 3 (Coveware): quarterly reports based on real IR engagements where Coveware acted as the negotiator. Payment-rate figures are representative of mid-market victims; very small businesses and very large enterprises are underrepresented in both directions.

Victim-as-perpetrator section: based on DOJ indictments and press releases, ProPublica investigative reporting, CISA advisories, the FATF March 2023 report on countering ransomware financing, and FinCEN's December 2025 Financial Trend Analysis. Cases listed are those that are publicly charged or plea-completed; ongoing investigations are not named.

What we did not do: assign attribution to any individual Icelandic incident, name any victim still under NDA, or suggest that any specific IR firm is compromised. The DigitalMint/Sygnia indictments are public facts — the pattern they represent is an industry-wide risk, not an accusation against any other party.

← All analysis Related: Is AI speeding up cyber attacks? Back to News Email the author