Daily brief at 7am Melbourne. Unsubscribe any time.

Sunday 26 July 2026

Cl0p Has a New Trick: Chaining Two PLM Bugs to Extort Manufacturers Without Ransomware

Cl0p is back with a fresh supply-chain play, SourTrade teaches browsers to assemble their own malware, and an unpatched Fastjson RCE is already being weaponised.

Lead story

Cl0p Has a New Trick: Chaining Two PLM Bugs to Extort Manufacturers Without Ransomware

The Cl0p crew — variously tracked as FIN11, Lace Tempest, and Graceful Spider — isn't content to keep repeating the MOVEit playbook. According to fresh reporting, affiliates linked to the group are now targeting internet-exposed deployments of PTC Windchill and FlexPLM, a pair of product lifecycle management platforms used heavily by manufacturers, defence contractors, and consumer goods companies worldwide.

The attack chain is elegant in a deeply annoying way. Attackers first hit a pre-authentication information-disclosure flaw in FlexPLM's WSDL endpoint — essentially a configuration file that leaks more than it should. They feed what they learn into a second vulnerability in the Windchill login servlet, which then lets them execute code on the server without ever providing valid credentials. No phishing. No brute-forcing. Just two bugs strung together into one quiet, automatic entry point.

What makes this especially notable is the pivot away from ransomware toward pure data extortion. Cl0p isn't encrypting files and demanding a decryption key — it's exfiltrating engineering data, supply-chain records, and design files, then threatening to publish them. This is the same model the group used after the MOVEit and GoAnywhere campaigns, and it works precisely because PLM systems sit at the heart of a manufacturer's crown jewels. The stolen data is often far more sensitive — and far harder to replace — than whatever a ransomware payment might recover.

Security researchers note that both Windchill and FlexPLM are commonly exposed to the internet for supplier and partner access, which is exactly the attack surface Cl0p is exploiting. That exposure is a feature for operations teams and a gift for threat actors.

Why it matters beyond the immediate victims. PTC's platforms are deployed across aerospace, automotive, industrial equipment, and defence manufacturing. Any organisation in those sectors running an unpatched, internet-facing deployment should treat this as a five-alarm situation. Shodan-style discovery of exposed instances is trivially easy for a well-resourced affiliate network.

For Australian readers, this warrants particular attention. Australia's defence industry supply chain — including tier-2 and tier-3 contractors supporting the AUKUS submarine programme, major infrastructure projects, and defence platform upgrades — relies heavily on PLM software from exactly this class of vendor. The SOCI Act's supply-chain risk obligations mean that any Australian critical infrastructure entity with an affected supplier should be checking whether those suppliers run exposed Windchill or FlexPLM instances, not just their own internal deployments. The ASD's Essential Eight wouldn't directly block the initial vector here, but patching and network segmentation of externally-facing PLM systems would.

What to watch. Cl0p has a well-documented pattern: discover a zero-day (or near-zero-day) in a widely-deployed file-transfer or collaboration platform, quietly exploit it at scale, then surface the victim list weeks later. The victims here may not know they've been hit yet. Expect a public extortion roster to emerge in the coming weeks.

In the meantime, if your organisation — or a supplier — runs either platform, the immediate action is straightforward: take the instance off the public internet if you can, apply available patches, and start reviewing outbound data transfer logs for anything anomalous.

Also today

SourTrade Makes Your Browser Build Its Own Malware

A malvertising campaign called SourTrade has been operating since late 2024 with an unusually clever evasion technique: rather than delivering a complete malicious executable from a fixed URL, it ships the payload in fragments and has the victim's browser assemble them into a Windows binary at runtime, using Bun — a legitimate JavaScript runtime — as its execution vehicle. The campaign has impersonated TradingView, Solana, and Luno to target retail traders and crypto users. Because no single request carries the full malicious file, traditional URL-based detection largely misses it. Security firm Confiant published its analysis on 23 July.

The Hacker News

Fastjson RCE Is Being Actively Exploited — and There's No Patch

Attackers are actively exploiting CVE-2026-16723, a CVSS 9.0 remote code execution flaw in Fastjson 1.x — Alibaba's widely-used JSON library for Java. In vulnerable Spring Boot applications, a crafted JSON request can execute arbitrary code without authentication, running at the privilege level of the Java process. Security firms ThreatBook and Imperva confirmed active exploitation. There is currently no patch available. Fastjson is embedded in a large number of enterprise Java applications globally, and its presence in Australian government and financial-sector software stacks makes this particularly urgent to triage.

The Hacker News

GitLab RCE PoC Published — Authenticated Users Can Run Commands as Git

Researchers at depthfirst have published a working exploit for a GitLab remote code execution vulnerability that GitLab patched back on 10 June. The flaw affects self-managed GitLab 18.11.3 servers that haven't applied the update. Any authenticated user with push access to a project can trigger it: commit a specially crafted Jupyter notebook, open its commit diff, and a heap memory leak cascades into code execution as the git system user. Six weeks is plenty of time for administrators to have patched — but many self-managed instances lag significantly on updates. If your organisation runs self-managed GitLab, this is a check-your-version-now situation.

The Hacker News

Insurance Phishing Has Upgraded to Real-Time Account Takeover

The old phishing model — harvest credentials, log in later — is giving way to something more immediate. CTM360's latest research documents insurance-focused phishing operations that intercept sessions in real time, relaying victims' credentials and MFA codes to live attacker sessions the moment they're entered. The victim thinks they're logging into their insurer's portal; the attacker is simultaneously authenticating on the real site in another window. The technique, sometimes called adversary-in-the-middle phishing, is not new in banking, but its spread to insurance — where accounts hold sensitive health, property, and life data — is a meaningful escalation.

The Hacker News

DevMan RaaS Offers Affiliates a Full Business Dashboard for Running Ransomware

PRODAFT has published research on DevMan, a ransomware-as-a-service operation it tracks as Funky Mantis, which provides affiliates with a slick web portal covering everything from payload customisation to victim management to affiliate payouts. The platform's level of polish — automated build generation, financial dashboards, integrated negotiation tools — reflects just how industrialised the ransomware economy has become. The operators appear to have invested meaningfully in usability, lowering the technical bar for affiliates and accelerating the pace at which new operators can spin up campaigns. Think Shopify, but for extortion.

The Hacker News

Anthropic's Opus 5 Bets on Efficiency Over Raw Power

Anthropic's new Claude Opus 5 doesn't try to leapfrog rivals on benchmark scores. Instead, it's engineered around token efficiency — delivering comparable output quality at significantly lower compute cost per task. That's a deliberate strategic choice: as frontier model capabilities converge, the competitive battlefield is shifting to price-per-useful-output rather than raw capability. Ars Technica's analysis notes that cheaper, smaller models are already good enough for most production workloads, and Opus 5 seems designed to defend Anthropic's enterprise position against that downward price pressure rather than chase another headline-grabbing capability milestone.

Ars Technica

One Fallen Power Line Revealed How Fragile AI Data Centres Really Are

A single downed power line in Northern Virginia — the world's most concentrated AI data centre corridor — exposed a systemic problem: most hyperscale facilities are not designed to gracefully handle sudden grid disruptions. When utility power dropped unexpectedly, generator failover sequences created transient voltage conditions that took several facilities offline simultaneously rather than isolating gracefully. TechCrunch's analysis examines the engineering fixes available, from improved automatic transfer switch configurations to smarter load-shedding protocols. The broader lesson is that AI infrastructure build-out has dramatically outpaced grid resilience planning — a risk that Australian operators scaling data centre capacity in Western Sydney and Melbourne face as well.

TechCrunch

Reid Hoffman and Mark Pincus Are Raising $100M to Automate Computer Tasks

Prentis, a new AI lab co-founded by LinkedIn's Reid Hoffman and Zynga's Mark Pincus, is reportedly in discussions to raise $100 million in early funding. The lab's central thesis is that automating routine computer tasks — navigating interfaces, filling forms, processing documents — will soon represent a larger commercial opportunity than coding assistance, which has dominated the first wave of enterprise AI adoption. It's a bet on the agentic layer becoming the primary value-capture point in enterprise software, above the model layer and below the application layer. How that plays out in a market where every major lab is also building agent frameworks will be interesting.

TechCrunch

Librarians Are Running Sold-Out 'Avoiding AI' Workshops

Public libraries across the United States are hosting "Avoiding AI" workshops — and they're filling up faster than almost any programming those libraries have offered before. The workshops teach participants how to identify AI-generated content, opt out of data collection by major platforms, and use non-AI-dependent tools for everyday tasks. The demand signals something the tech industry tends to underweight: a substantial portion of the public is not enthusiastic about AI's ubiquity and actively wants practical tools to minimise exposure. Librarians, with their tradition of digital literacy and privacy advocacy, have quietly become the most credible voice in the AI-sceptic space.

TechCrunch

Big Tech Is Stonewalling European Researchers Entitled to Its Data

Researchers studying social media's societal effects say TikTok, X, and Meta are not providing the platform data they are legally required to share under the EU's Digital Services Act — the regulation specifically designed to give academics access to algorithmic and content-moderation data. The accusations range from throttled API access to outright non-compliance with data-sharing requests. The DSA has enforcement teeth, but regulators have been slow to use them. Australia's Online Safety Act contains analogous transparency provisions, and the ACMA has signalled similar frustrations with platform responsiveness — making this a genuinely parallel story on both sides of the globe.

Ars Technica

Vietnam Moves Toward Kids' Social Media Restrictions as the Global Trend Accelerates

Vietnam is drafting legislation to restrict social media access for minors, joining a growing list of countries following Australia's lead. Australia was the first nation to legislate an outright ban in late 2025, targeting children under 16 on major platforms. Since then, the policy has been debated or advanced in the EU, UK, US states, and now Southeast Asia. TechCrunch's tracker documents the spread of the trend and the varying regulatory approaches — age verification mandates, outright bans, design-duty requirements. Australia's experience — including the enforcement challenges and platform pushback — is increasingly being studied as an early-mover case study.

TechCrunch

Sources consulted