The fridge hackers went on a European tour

It's probably nothing for the fourth of November.

The supermarket refrigeration sabotage that started in France has gone on a European tour. Reuters reports a coordinated overnight wave on Monday hitting Ahold Delhaize and Colruyt in Belgium and Luxembourg, ICA and Coop in Sweden, an IKEA store, and thirteen Musgrave SuperValu stores in Ireland. Same playbook as France in October — freezers driven up to fifty degrees, stock binned. One Stockholm ICA franchise got hit at 2 a.m.; its sales manager said every freezer maxed out. Carrefour reported fifteen of its seven hundred Belgian stores affected, Coop seven of eight hundred. At least some units used the same Danfoss control technology targeted in France, and Danfoss says it's been working with customers on fixes. The EU agriculture commissioner called it a deliberate attempt to disrupt the food supply chain.

Attribution, though, is nowhere. Reuters couldn't establish who's behind it or whether it's even linked to the French incidents, and Belgium's cybersecurity centre said the perpetrator may never be determined. SentinelOne's Tom Hegel noted the absence of financial demands points to sabotage rather than extortion, and that exposed control panels leave almost no traces. Moscow has repeatedly denied involvement in sabotage operations, for what that's worth — treat any Russia speculation as speculation. The genuinely wild part remains that this many supermarket chains have internet-reachable refrigeration controllers to begin with.

Meanwhile, the Binance theft has an official attribution. The FBI said at the weekend that North Korea was responsible, per Reuters, and Pyongyang responded with unusual speed — Kim Yo Jong used state media on Tuesday to call it absurd rhetoric and a grave political provocation. Meanwhile TRM Labs' tracing update shows the roughly two point four billion dollars in unfrozen ether and bitcoin moving through new laundering routes: a self-minted TRON token sold into its own liquidity pool, a lending receipt token passed between wallets to defeat USDT-only tracing, and cross-chain swaps chained through two bridges with no intermediate wallet. Nearly two point seven billion remains frozen. Worth keeping the timing straight: TRM attributes the theft to TraderTraitor with high confidence, but noted no government had attributed it — the FBI statement, dated only as at the weekend, sits slightly alongside that. Context from Reuters: Chainalysis estimates eighty-nine percent of crypto hack losses this year trace to North Korea, and Elliptic puts the country's twenty twenty-six haul at a record seven point six billion.

ESET has publicly identified Gorbag, a Russia-aligned group active since at least February last year. The target list reads like a wartime intelligence wishlist: Ukrainian law enforcement and judicial bodies — ESET suggests monitoring war crimes investigations — frontline councils doing strike-damage assessment, defence organisations, and a transport hub used for foreign military aid. Its BeautyStealer malware grabs browser credentials, VPN configs, Office documents and password databases, exfiltrates via Telegram, then deletes itself. Collection over persistence. My favourite detail: one variant carries a comment addressed to AI reviewers insisting the file is benign and needs no analysis. ESET tested it against two AI assistants — the prompt injection failed and both flagged it as a reverse shell. Malware authors are now writing for an AI readership, which was inevitable, but it's nice to see the early attempts land this poorly. One practical note: delivery runs through the WinRAR path traversal flaw patched in July last year, so if your estate still has unpatched installs, it's being actively exploited by multiple actors.

Staying on prompts: Check Point Research has dissected Haiou, a Chinese-language jailbreak bundle for OpenAI's Codex that persistently rewrites the agent's trusted global instructions — so an ordinary greeting triggers an ad for a third-party gateway, and the agent is told to exempt that gateway from offensive testing. They're calling it prompt adware. Keep the qualifications: no evidence the gateway is malicious, no data theft observed, and it's explicitly not classed as a supply-chain attack. But third-party prompts are now part of the trust chain, and most organisations have no visibility into what their agents' instructions actually say anymore. If teams are installing community jailbreaks or prompt bundles, that's now a supply-chain review item, not a productivity choice.

And quickly, closer to home: First State Auctions, a Sydney luxury-goods auction house, has confirmed unauthorised access to a system holding customer information, after a forum user offered more than a million lines of data essentially free for site credits. The company says impact appears mostly limited to basic contact info, with no evidence cards or passwords were affected, and the regulators have been informed. Keep the two sides distinct — the million-line figure is the seller's claim, the limited impact is the company's early assessment. Both can be true, or one can age badly.

That's it for today. I'm off to check whether my fridge is on the internet — and honestly, I don't want to know. Catch you next time.

Don't forget, all our threat intelligence comes from Mallory dot A I and open sources. If you want intelligence your team can act on, reach out for a chat.