Tuesday 1 September 2026

ShinyHunters Claims 284 Million McKesson Records — and the Clock Is Already Ticking

McKesson confirms a 284-million-record breach with a ransom clock ticking, China's Fire Ant pivots from VMware to Cisco routers, ChatGPT Ads hits $1B, and North Korean job fraudsters infiltrate healthcare.

Lead story

ShinyHunters Claims 284 Million McKesson Records — and the Clock Is Already Ticking

Healthcare and pharmaceutical logistics giant McKesson has confirmed a cyberattack on its systems, with the notorious ShinyHunters extortion group claiming it swiped 284 million records and threatening to publish everything if a ransom isn't paid. The company has brought in outside investigators and told regulators it's still in the early stages of understanding the damage — which, at this scale, could take a while.

McKesson is not a household name to most people, but it is the circulatory system of US healthcare. The company moves roughly one-third of all pharmaceuticals distributed in the United States and supplies hospitals, pharmacies, and healthcare networks across North America and beyond. A breach of this magnitude doesn't just mean leaked email addresses — it potentially means patient data, prescription records, and supply chain intelligence sitting in criminal hands.

ShinyHunters' track record makes this credible. The group has previously claimed and delivered on large-scale data thefts from Ticketmaster, Santander Bank, and dozens of others. Their method is typically to post a sample, set a deadline, and follow through on publication if payment isn't made. McKesson has not confirmed exactly what data was taken or whether it originated from a third-party application — the company's own filing describes "an unnamed third-party application" as the entry point, which raises its own questions about supply chain risk management.

The timing is awkward. McKesson is currently warning of "service degradation," which suggests operational impact beyond just data exfiltration. Healthcare logistics disruption ripples fast — delayed pharmaceutical deliveries are a patient safety issue, not just a business continuity problem.

For Australian readers, the parallel risk is real. McKesson has distribution operations in Australia through its Australian Healthcare Solutions arm, and a number of Australian hospital networks and pharmacy groups are part of global pharmaceutical supply chains that intersect with companies like McKesson. The SOCI Act's third-party risk requirements exist precisely for scenarios like this — a critical infrastructure provider compromised via an upstream software dependency. If your organisation has any supply chain exposure to McKesson or its technology partners, now is the time to ask questions.

The attacker's deadline is not yet publicly confirmed, but ShinyHunters historically moves quickly. Defenders in adjacent industries — anyone in pharma logistics, hospital networks, or healthcare IT — should be monitoring for data dumps and preparing customer notification playbooks.

What to watch: whether McKesson identifies the breached third-party application (which would tell us a great deal about the attack surface), whether data actually appears online when the deadline hits, and whether US regulators treat this as a HIPAA mass-notification event. A 284-million-record claim from a group with ShinyHunters' batting average deserves to be treated as credible until proven otherwise.

Also today

China's Fire Ant Pivots from VMware to Cisco Routers in Espionage Campaign

Incident response firm Sygnia has detailed how a China-linked espionage group, tracked as Fire Ant, has expanded beyond VMware hypervisors to compromise Cisco IOS XR routers, TACACS authentication servers, and Linux management hosts. The group's goal appears to be positioning itself deep inside high-value network infrastructure — stealing credentials, erasing logs, and establishing persistent access that's hard to detect and harder to evict. The expansion to routing infrastructure is significant: routers are the nervous system of enterprise networks, and compromising TACACS means owning authentication for everything behind it. Organisations running Cisco IOS XR in sensitive environments should treat this as a priority review item.

The Hacker News ↗

North Korea's Fake IT Worker Scheme Spreads Into Healthcare and Sales Roles

North Korea's long-running IT worker fraud scheme — where operatives pose as legitimate remote employees to earn hard currency and steal data — has now extended beyond tech roles into healthcare and sales and marketing positions. The expansion matters because healthcare roles often carry access to patient data and clinical systems, while sales roles can expose CRM databases and deal pipelines. The scheme has already cost Western companies tens of millions in lost data and sanctions violations. Australian employers hiring remote contractors in these fields should note that the Australian Cyber Security Centre has previously flagged DPRK IT worker activity as a live threat to Australian businesses, particularly in fintech and professional services.

The Hacker News ↗

Prompt Injection Hidden Inside a Court Filing — and It Worked

Bruce Schneier flagged a novel and genuinely alarming case: someone embedded AI instructions inside a legal document filed with a court, designed to manipulate any AI tool that processed the filing. If a lawyer, clerk, or judge used an AI assistant to summarise or analyse the document, the hidden instructions could redirect the AI's output. This is a textbook prompt injection attack — the AI equivalent of a SQL injection — but executed in a context most people haven't considered: legal proceedings. As AI-assisted document review becomes standard practice in law firms and courts, adversarial documents become a real attack vector.

Schneier on Security ↗

ATM Vulnerabilities Expose Deeper Software Supply Chain Rot

A security researcher disclosed nine vulnerabilities in encryption and authentication software used across ATM networks. The flaws affect not just the machines themselves but the underlying software stack that handles authentication and cryptographic operations — meaning the problem extends to any system sharing that codebase. It's a useful reminder that ATMs are not bespoke banking hardware; they're Windows PCs running legacy software stacks with decades of inherited dependencies. Five people pleaded guilty this week in a separate US federal ATM jackpotting case, suggesting physical and digital ATM attacks are both alive and well in 2026.

WIRED Security ↗

ChatGPT Ads Hits $1 Billion Annualised Revenue — and Expands Globally

OpenAI announced its ChatGPT Ads product has reached a $1 billion annualised revenue run rate and is expanding to more markets. The pitch is that ad-supported ChatGPT access funds free and low-cost AI for users who can't pay subscriptions. It's a classic two-sided market play — advertisers fund access, users trade attention for capability. Whether it changes ChatGPT's behaviour or editorial independence as an information source is an open question that regulators in the EU and Australia's ACMA will likely be watching closely, given both jurisdictions have live debates about AI and online content accountability.

OpenAI Blog ↗

ChatGPT and Reddit Now Subject to the EU's Toughest Online Safety Rules

Explosive user growth has pushed both ChatGPT and Reddit over the threshold that triggers designation under the EU's Digital Services Act as Very Large Online Platforms, meaning they now face the bloc's most demanding content moderation, transparency, and risk assessment obligations. For OpenAI, this arrives at the same time as its advertising expansion — a combination that will draw regulatory scrutiny about how AI-generated content and paid placements interact. Australia's Online Safety Act has similar (if less stringent) large platform obligations, and the eSafety Commissioner has signalled interest in how AI platforms handle harmful content at scale.

Ars Technica ↗

Nvidia Bets $3.5 Billion on MediaTek to Stay Relevant as Big Tech Builds Its Own Chips

Nvidia is investing $3.5 billion into Taiwanese chipmaker MediaTek in a deal that signals Nvidia's strategy for the era when hyperscalers build their own AI silicon. Rather than fighting Google's TPUs or Amazon's Trainium head-on, Nvidia wants to be the architecture and IP backbone that even custom chip builders rely on. MediaTek's reach into edge devices, smartphones, and automotive systems gives Nvidia a foothold in markets its data-centre GPUs don't easily address. For Australia, Nvidia's chipmaking partnerships directly affect the cost and availability of AI inference hardware that local cloud providers and government agencies procure.

TechCrunch AI ↗

TerminalFix: The PowerShell Campaign Using PNG Files to Hide Its Tracks

A new enterprise-targeted attack campaign dubbed TerminalFix is running a sophisticated multi-stage chain that hides malware payloads inside PNG image files and deploys a custom reverse tunnel into victim networks — giving attackers persistent, stealthy access that blends into normal HTTPS traffic. The initial lure is a ClickFix-style social engineering trick: users are convinced to run a PowerShell command that kicks off the entire chain. The reverse tunnel component is what elevates this above commodity malware — it allows attackers to maintain access even through NAT and firewall boundaries without opening inbound ports.

Dark Reading ↗

The Pentagon Gets Its Own ChatGPT and Grok

The US Department of Defense has added OpenAI's ChatGPT and SpaceX AI's Grok to its centralised AI portal, joining Google's Gemini. The move represents a significant step toward institutionalising commercial large language models within defence operations, with the Pentagon presumably running these on isolated or controlled infrastructure rather than consumer endpoints. The decision to include Grok — Elon Musk's AI product — inside the US military's official toolset will raise eyebrows given ongoing debates about Musk's access to sensitive government systems. It also follows a court ruling this week that found the Pentagon's earlier actions against Anthropic — labelling it a supply chain risk — were illegal.

TechCrunch AI ↗

Check Point Unpacks JSCeal: The Stealer That Hides in Node.js Bytecode

Check Point Research has published a detailed reverse-engineering walkthrough of JSCeal, a credential-stealing malware that hides inside compiled V8 bytecode — the same format used by Node.js applications. Because most security tools scan JavaScript source, not compiled bytecode, JSCeal slips past defences that would catch conventional infostealers. The malware specifically targets cryptocurrency wallets and applications. Check Point's analysis demonstrates a static deobfuscation technique that can unpack the bytecode for analysis — a useful contribution for defenders writing detection rules. The malware has been active since March 2024 and is also tracked as WEEVILPROXY and MeadowLocust.

Check Point Research ↗

Instagram Clamps Down on Undisclosed AI Profiles

Meta has introduced new restrictions on AI-generated profiles on Instagram that don't clearly identify themselves as artificial. Accounts that aren't labelled as AI will face reduced reach and distribution, effectively penalising undisclosed bot accounts without removing them outright. The move follows growing user frustration with AI influencers — generated personas that post content, build followings, and in some cases promote products without ever disclosing their non-human nature. Australia's eSafety Commissioner and the ACCC's ongoing digital platform inquiry have both flagged AI-generated deceptive content as a priority concern, so Meta's self-regulatory move here may forestall harder regulatory intervention in this market.

TechCrunch AI ↗

Sources consulted