Daily brief at 7am Melbourne. Unsubscribe any time.

Tuesday 13 October 2026

Ukraine Hits a Third Yandex Data Centre, and Russia's Internet Is Starting to Crack

Ukraine's drone strikes on Yandex data centres have taken Russia's digital economy to its knees — and exposed just how dangerously centralised it was built.

Lead story

Ukraine Hits a Third Yandex Data Centre, and Russia's Internet Is Starting to Crack

Over the weekend, Ukrainian drones struck a third Yandex data centre in Russia, forcing it offline and pushing the country's internet infrastructure into genuine crisis. Cloud connectivity reportedly fell to around 30% of normal. Payments stalled. Deliveries ground to a halt. For millions of Russians, the digital services they rely on every day simply stopped working.

This isn't just a dramatic headline about servers on fire. It's a case study in what happens when a country builds a hyper-centralised digital economy and then goes to war with a neighbour that has drone capabilities.

The structural problem Russia is now paying for

Yandex — often described as "Russia's Google" — is not just a search engine. It operates cloud infrastructure, mapping, food delivery, ride-sharing, fintech, and more. Russia's internet economy consolidated heavily around a small number of domestic platforms after the country began forcing tech companies to store data locally and severed ties with Western cloud providers. That strategy was meant to create sovereignty. Instead, it created a handful of extraordinarily high-value targets.

Russian bloggers have been sounding the alarm about this vulnerability publicly, warning that the country's hyper-centralised digital economy left it exposed to exactly this kind of attack. Those warnings now look prescient.

The bot traffic debate

A Cloudflare chart circulating on social media appeared to show bot traffic from Russia collapsing in the wake of the attacks. The interpretation that "Ukrainian drones took out Russia's bot farms" spread quickly. The reality is more nuanced: the bots most affected appear to be commercial crawlers and automated web services running on Yandex infrastructure — not the kind of geopolitical influence-operation bots most people had in mind. Still, the disruption to legitimate digital services is significant enough on its own.

What Yandex's engineers are dealing with

Engineers have reportedly been working around the clock to reroute traffic, restore services, and compensate for lost capacity. The repeated nature of the strikes — three data centres now — suggests Ukraine is executing a deliberate campaign rather than targeting targets of opportunity. Each successive hit compounds the recovery difficulty.

Why this matters beyond the conflict

For security and infrastructure planners outside Russia, this is a live demonstration of a threat scenario that's been theorised for years: adversarial drone strikes on civilian digital infrastructure in a peer conflict. Physical security of data centres — not just logical security — is now a serious operational concern.

It also raises questions about concentration risk that apply everywhere. Australia's own Critical Infrastructure Act and SOCI Act frameworks require operators to assess physical security and single-point-of-failure risks. Watching Russia's experience in real time is about as vivid a reminder as the threat-modelling literature is ever going to produce.

What to watch

Whether Yandex can restore meaningful service in the coming days will test the resilience of whatever redundancy it had left. More broadly, watch for whether other Russian digital infrastructure — beyond Yandex — becomes a target. If this campaign widens, the disruption could become existential for Russia's domestic internet economy in ways that no sanctions package has managed.

Also today

Cyberattack Knocks Out Refrigeration at Hundreds of French Supermarkets

A cyberattack overnight Sunday disrupted refrigeration systems across hundreds of French supermarkets, spoiling perishable food at stores run by some of the country's largest grocery chains. The incident is a sharp reminder that operational technology (OT) — the systems that control physical equipment like chillers and HVAC — remains a soft target. Unlike IT breaches that leak data, OT attacks cause immediate, tangible harm: wasted food, supply chain disruption, and potential food safety risks. Australian supermarket operators and cold-chain logistics providers subject to SOCI Act obligations should be paying close attention — retail OT security has lagged well behind IT for years.

The Record ↗

Ledger Finds Unauthorised Hardware Implant in Devices Sold by Reseller CryptoBilis

Ledger has paused sales through Southeast Asian reseller CryptoBilis after discovering an unauthorised hardware implant in devices linked to more than $86 million in stolen funds. Hardware supply chain attacks on crypto wallets are particularly insidious — the device looks legitimate, passes basic inspection, and may function normally until it silently exfiltrates seed phrases. Ledger's official advice remains to buy only through its own store or authorised retailers, and to never enter a seed phrase into any device that arrived pre-configured or in a tampered package. With crypto adoption rising in Australia, users should be especially wary of wallet purchases through third-party marketplaces.

The Hacker News ↗

CVSS 10.0 SonicWall Flaw Already Being Exploited Days After Disclosure

A maximum-severity pre-authentication server-side request forgery (SSRF) vulnerability in SonicWall's Secure Mobile Access 1000 appliances — tracked as CVE-2026-102255 — is already seeing active exploitation, with 135 attempts logged within days of the patch being published. SSRF flaws let attackers force a vulnerable device to make requests on their behalf, which can be used to probe internal networks or pivot further into an environment. SonicWall appliances are widely deployed as VPN and remote access gateways in Australian enterprises and government agencies; if you run SMA 1000 series hardware, patching this is not optional.

SecurityWeek ↗

Bricksforge WordPress Plugin Hit by Unauthenticated RCE Attacks

A CVSS 10.0 flaw in the Bricksforge plugin for WordPress is being actively exploited in the wild. The vulnerability, CVE-2026-85097, allows completely unauthenticated attackers to upload and execute arbitrary PHP code — effectively handing them full control of a vulnerable site. Bricksforge is a popular page-builder extension, meaning a wide range of sites are potentially at risk. If you manage WordPress sites running Bricksforge, update immediately. With WordPress powering a large proportion of Australian small-business and government websites, this one deserves urgent attention from web administrators.

The Hacker News ↗

Meta Launches Muse Wonder: Its First Frontier-Class AI Model

Meta has released Muse Wonder and Muse Wonder Full, its first frontier-scale AI models from the revamped Superintelligence Labs division. Unlike the lightweight Muse Spark, Wonder targets high-complexity tasks including coding and reasoning, and posts benchmark scores the company claims are competitive with the industry's best. Wonder is available now to Muse Code subscribers and via the Meta Model API. Meta plans to eventually power its Muse AI agent with Wonder. For enterprises evaluating frontier models, this meaningfully expands the competitive field beyond OpenAI, Anthropic, and Google — and Meta's open-weight track record could matter for Australian organisations with on-premise requirements.

TechCrunch AI ↗

Microsoft's Maia 300 Chip Aims to Reduce Reliance on Nvidia GPUs

Microsoft has announced its third-generation in-house AI chip, Maia 300, designed to serve as the primary inference system across its products and services. The chip is optimised for efficiency and is built to handle the latest reasoning and agentic AI workloads. Production rollout is scheduled to begin in 2027. The strategic intent is clear: reduce dependence on Nvidia's expensive GPUs for running AI at scale in Microsoft's data centres. For CIOs evaluating Azure as an AI platform, this points to lower long-term inference costs — and for Nvidia investors, it's another reminder that hyperscalers are serious about custom silicon.

The Verge ↗

CyberXero Used Claude Code to Steal 628,000 Ukrainian Records

Researchers have detailed a campaign by threat actor CyberXero that combined commodity tools like Cobalt Strike with Anthropic's Claude Code as an AI orchestration layer to attack WordPress sites, e-commerce platforms, and critical infrastructure in Ukraine. The result: more than 628,000 records stolen. The campaign illustrates how AI coding assistants are lowering the skill floor for sophisticated attacks — not by replacing human hackers entirely, but by accelerating the parts that used to require specialist knowledge. It also puts Anthropic in an uncomfortable position; Claude Code was designed for legitimate developers, not offensive campaigns.

The Hacker News ↗

Bring Your Own Agent: The Shadow IT Problem Nobody Budgeted For

Personal AI agents — the kind employees are quietly connecting to corporate accounts without IT's knowledge — are becoming a serious enterprise security headache. Amazon has already blocked Meta's Muse agent because it doesn't identify itself as a bot while browsing. Privacy concerns are piling up around others. The pattern mirrors the early days of shadow IT and BYOD, except this time the unauthorised tool isn't a personal Dropbox account — it's an autonomous agent with access to email, calendars, and internal systems. Organisations without an AI agent usage policy are arguably already behind. Australian privacy obligations under the Privacy Act 1988 make this a compliance issue as much as a security one.

The Hacker News ↗

IonQ to Acquire Zero-Trust Firm Elisity for Up to $1.3 Billion

Quantum computing company IonQ is acquiring zero-trust security firm Elisity in an all-stock deal valued at up to $1.3 billion. The combination pairs Elisity's least-privilege network access controls with IonQ's post-quantum cryptography capabilities — a strategic bet that enterprises accelerating toward quantum-resistant encryption will want both in one place. Federal agencies and large corporates are the obvious target market as government timelines for quantum-ready infrastructure tighten. It's an unusual pairing, but the underlying logic is sound: post-quantum crypto and modern access controls are two halves of the same migration problem.

CyberScoop ↗

Let's Encrypt Shortens Default Certificate Validity to 64 Days

Let's Encrypt will cut its default TLS certificate validity period to 64 days beginning February 2027, as a stepping stone toward the industry's planned 45-day maximum. Shorter-lived certificates reduce the window an attacker has to exploit a compromised or misissued cert — a relevant concern given recent TLS trust incidents. The practical implication: any organisation still managing certificates manually needs to automate renewal before this change lands. For Australian organisations using Let's Encrypt certificates — common in government and SME web infrastructure — failing to automate will mean regular outages, not just a security problem.

SecurityWeek ↗

AI Tool 'Raven' Lets Police Identify Properties From a Single Interior Photo

A company called Raven — formerly GeoSpy — has built an AI tool that lets law enforcement identify which specific Airbnb listing or property a photo was taken in, matching interior images against listing databases to pinpoint locations. Separately, court documents reveal US federal investigators obtained warrants to deploy fake cell-site simulators (IMSI catchers) against a journalist and a political candidate. Together, these stories paint a picture of AI dramatically expanding law enforcement's physical surveillance capabilities. Australia has no equivalent of the US Fourth Amendment, and the use of similar AI-assisted location tools by Australian police would face scrutiny under the Privacy Act and relevant surveillance device laws.

404 Media ↗

Sources consulted