Daily brief at 7am Melbourne. Unsubscribe any time.

Monday 7 September 2026

Australia's Singapore Cable Is Broken — and It's Not the First Time

A submarine cable break in Indonesian waters has cut Vocus's Australia-Singapore link — a reminder that undersea infrastructure is both critical and fragile.

Lead story

Australia's Singapore Cable Is Broken — and It's Not the First Time

A break in the Australia Singapore Cable (ASC) in Indonesian waters has knocked out one of Vocus's key international links, the telco confirmed early Monday morning. The break joins a growing list of subsea cable incidents that have quietly shaped how Australians — and Australian businesses — connect to the world.

Vocus hasn't confirmed how much traffic has been disrupted or whether it's been rerouted, but the company operates the ASC as a critical artery for data moving between Australia, Singapore, and onward to Europe, the Middle East, and Asia. For enterprises relying on Vocus for international connectivity, this is the kind of event that stress-tests redundancy plans you hope you never need.

Why this matters more than a typical outage

Submarine cables aren't just a telco problem — they're national infrastructure. Australia's internet connectivity depends on a relatively small number of physical cable systems. When one breaks, traffic has to be rerouted across the remaining links, which increases latency and, in congested periods, can meaningfully degrade service.

The ASC runs roughly 4,600 km and connects Perth to Singapore. Indonesia's territorial waters sit squarely in its path, and that stretch is one of the more exposed sections — fishing activity, vessel anchoring, and seismic events all pose ongoing risks to cables in the region.

This is also not a one-off. The broader Asia-Pacific region has seen a cluster of subsea cable incidents in recent years — some accidental, some deliberately targeted. In 2022 and 2023, cables serving Tonga and Pacific Island nations were cut, leaving communities effectively isolated. Taiwan's cables have been repeatedly severed in incidents that US officials have described as deliberate. The question of how much is accident and how much is something else has never fully been resolved.

The strategic picture

Australia's Critical Infrastructure (SOCI) Act designates submarine cables as critical infrastructure, and the government has expanded its powers to respond to incidents affecting them. The Australian Communications and Media Authority (ACMA) licences submarine cable landing stations, and the government has previously blocked foreign involvement in cable projects citing national security concerns.

But regulation and physical protection are different things. A cable sitting on the ocean floor in international or Indonesian waters is very difficult to guard. Repair ships — there are only around 60 in the world — are booked out months in advance. A repair timeline of weeks to months is not unusual once a fault is located and a ship dispatched.

What to watch

Vocus has not yet indicated a repair timeline or the cause of the break. If rerouting is absorbing the traffic cleanly, business customers may not notice much. If the remaining capacity is strained, enterprise customers and cloud providers with AU-Singapore traffic flows are the most likely to feel it first.

Longer term, this incident adds weight to the case for diversifying Australia's cable routes — something the federal government and private operators have discussed but where progress remains slow. A second cable connecting Australia's east coast directly to India, bypassing the Singapore chokepoint, has been floated but not funded.

For now, Vocus customers should be checking with their account teams about current routing status. And the rest of us should be reminded that the internet's backbone is, quite literally, a wire on the ocean floor.

Also today

MikroTik Routers Being Hijacked via Unauthenticated SSH — No Patch Yet

Poland's CERT has issued an urgent warning after attackers were found seizing full administrative control of MikroTik routers through internet-exposed SSH services — with no authentication required. Successful attacks have been confirmed dating back to at least 2 September. MikroTik devices are widely deployed by ISPs, enterprises, and small businesses globally, including across Australia. The attack vector is straightforward: if your router's SSH port is reachable from the internet, it's potentially already compromised. CERT Polska has not yet published a victim count, but the advice is unambiguous — disable internet-facing SSH on MikroTik devices immediately if it isn't strictly necessary, and audit logs for any unexplained admin activity.

The Hacker News ↗

REVSTEALER's Ghost Modules Kill Windows Defender Before Mining Crypto

Elastic Security Labs has documented four previously unknown programs that REVSTEALER — an emerging Windows infostealer — leaves behind after it finishes its job and deletes itself. The persistence modules, named ProManager, WinUpdate, SoftManager, and a fourth not yet publicly named, are designed to outlast the initial infection. Most alarmingly, one of them disables both Windows Update and Microsoft Defender before deploying a cryptocurrency miner. The technique is a reminder that cleaning an infostealer infection isn't enough — defenders need to verify that the underlying OS protections haven't been quietly switched off and that the system isn't doing something profitable for someone else in the background.

The Hacker News ↗

OpenAI Chief Scientist Calls AI an 'Alien Mind' — and Warns Safeguards Aren't Keeping Up

OpenAI's chief scientist Jakub Pachocki has published a striking essay describing increasingly capable AI systems as possessing a form of 'alien mind' — one that thinks differently from humans in ways that make alignment genuinely hard. He calls for stronger internal safeguards at labs and, notably, for international coordination on AI safety standards before capabilities outpace governance. The piece lands at an interesting moment: OpenAI has been publicly positioning itself as safety-conscious while simultaneously shipping increasingly autonomous agents. Pachocki's framing is unusually candid for a sitting lab executive, and his call for international coordination echoes what Australia's AI Safety Institute has also been pushing at a multilateral level.

OpenAI Blog ↗

Inside OpenAI: Coding Agents Are Already Reshaping How AI Research Gets Done

OpenAI has published internal data showing that AI coding agents are meaningfully accelerating the pace of research inside the company. Experiment velocity is up, tasks are growing in complexity, and researchers are offloading more of the grunt work of coding to agents. The piece is partly self-promotional, but the underlying data point is genuinely significant: if the world's most capable AI lab is using AI agents to build better AI, the feedback loop has already begun. For Australian AI researchers and R&D teams, the implication is that teams not experimenting with agent-assisted workflows are already at a structural disadvantage relative to labs running them at scale.

OpenAI Blog ↗

Authors vs. Publishers: The Anthropic Settlement Is Already Getting Messy

Authors who were part of the Anthropic copyright settlement are pushing back against publishers and literary agents who are claiming a significant share of the settlement payments. The authors' argument is that publishers are claiming more than they're entitled to given that it was the authors' creative work — not the publishers' investment — that was allegedly used to train the AI model. The dispute is a preview of a much larger set of questions the creative industries will face as AI copyright settlements become more common. Australia has its own unresolved questions around AI and copyright, with the Attorney-General's review of the Copyright Act still pending formal recommendations.

TechCrunch ↗

Seattle Times and Newsday Sue OpenAI and Microsoft Over Training Data

Two more American news publishers — the Seattle Times and Newsday — have filed suit against OpenAI and Microsoft, alleging their journalism was scraped and used to train AI models without authorisation or compensation. The lawsuits follow a growing wave of similar actions from the New York Times, Chicago Tribune group, and others. OpenAI and Microsoft have consistently disputed the characterisation, arguing their use falls within fair use doctrine. The flood of litigation is shaping up to be one of the defining legal battles of the AI era — one that Australian publishers are watching closely given that Australian copyright law doesn't have an equivalent fair use exception.

TechCrunch ↗

Travis Kalanick Is Back — This Time With a Robotaxi Startup

Travis Kalanick, who cofounded Uber and was pushed out as CEO in 2017, is reportedly pivoting his current venture Atoms toward the robotaxi business. Kalanick has described it as a chance to finish 'unfinished business' — a pointed reference to the autonomous vehicle ambitions Uber ultimately sold to Aurora. The robotaxi space is already crowded with Waymo, Tesla's Cybercab, and a cluster of Chinese operators pushing into international markets. Whether Kalanick's entry represents genuine competitive differentiation or nostalgia capital at scale remains to be seen, but his return to the space will draw attention and, almost certainly, investment.

TechCrunch ↗

Tesla's Cybercab Is on the Road — But Already Running Into Problems

Tesla's Cybercab robotaxi has entered limited road operations, but the rollout is already hitting snags. TechCrunch's mobility newsletter reports early operational issues, though full details weren't disclosed at time of publication. The Cybercab launch is a high-stakes moment for Tesla and Elon Musk, who has staked significant credibility on autonomous vehicles delivering returns after years of promises. Waymo's measured city-by-city expansion has set a cautious industry template; Tesla's approach of pushing deployment fast and iterating later is familiar from its earlier driver-assist rollouts — a strategy that has drawn both admirers and regulators.

TechCrunch ↗

Europe Has Its First Fully Commercial Orbital Rocket

German startup Isar Aerospace has successfully placed its two-stage Spectrum rocket into low Earth orbit, becoming the first entirely commercial European company to achieve orbital launch. The milestone comes after a failed March attempt that ended with the vehicle ditching into the sea 30 seconds after liftoff. This time, Spectrum launched from a Norwegian spaceport and completed the mission. Europe's launch capability has long been dominated by national programmes and the intergovernmental Ariane consortium; Isar's success signals that private capital can now compete in what was once exclusively government territory — a shift with implications for how Europe positions itself in the growing commercial space economy.

The Verge ↗

Amazon Cargo Plane Overruns Runway at Miami International

An Amazon-branded cargo aircraft operated by 21 Air overran the runway at Miami International Airport on Sunday, crashing into vehicles and injuring multiple people. The FAA confirmed the incident involved Flight 7598 and that an investigation is underway. The full extent of injuries and property damage was unclear at time of reporting. Amazon uses third-party cargo carriers extensively for its logistics network rather than operating its own aircraft directly, which complicates questions of liability and operational oversight. Aviation regulators will be scrutinising the carrier's maintenance records and approach procedures as part of the standard incident review.

The Verge ↗

The Fairphone 6 Plus Review: Ethical and Actually Good Now

The Verge's review of the Fairphone 6 Plus finds that the latest iteration has finally bridged the gap between sustainability mission and user experience. Previous Fairphone models required genuine compromise — slow processors, mediocre cameras — in exchange for ethical materials sourcing and excellent repairability. The 6 Plus, by contrast, feels like a competitive midrange Android phone that also happens to be repairable and made with responsibly sourced components. Right to repair has become a live policy debate in Australia, with the government's repair consultation ongoing. Fairphone's commercial success — if it holds — provides a useful data point that repairability doesn't have to mean sacrificing product quality.

The Verge ↗

Sources consulted