74,000 corporate firewalls cracked — the worst part is how ordinary the break-in was

Information Technology 4 min 80 sources

A self-feeding password attack hit half the internet's Fortinet firewalls, including ones at Oracle, Samsung and a NATO contractor. Plus AI splits the G7, Apple warns of price hikes, and the open web tries email.

Key takeaways

Attackers cracked login credentials on about 74,000 Fortinet firewalls — half the ones exposed online — using old leaked passwords, not a new flaw, and the haul includes Oracle, Samsung and a NATO contractor.
At the G7, world leaders pushed back on relying on American AI after Washington abruptly blocked exports of Anthropic's top models — convenience has revealed itself as a dependency that can be switched off.
Apple says price rises are now "unavoidable" because AI data centres are buying up the memory chips, pushing costs up fourfold in a year.

The biggest tech story today is not a clever new hack. It is an old one, run at a scale that should worry anyone who keeps a company’s front door on the internet. Researchers say attackers have cracked the login credentials of nearly 74,000 Fortinet firewalls — the boxes that sit at the edge of a corporate network and decide who gets in. That is roughly half of all Fortinet firewalls currently exposed to the internet [6].

The break-in that needed no genius

The campaign, named FortiBleed by the firms that found it, did not exploit a secret flaw [9]. The attackers scanned the internet for Fortinet devices, then tried passwords already known from past leaks. Where a company had never changed the password, they walked in [9].

Then it got worse. Once inside a firewall, the attackers used it as a listening post — watching the traffic flow through and scooping up any fresh credentials that passed by. Those new passwords fed the next round of break-ins. “The system feeds itself,” one of the firms, SOCRadar, wrote [9].

The list of compromised organisations is not small. Researchers named Oracle, Lenovo, Samsung, Comcast, Foxconn, PwC, Accenture, Siemens — and a NATO defence contractor [6][9]. The stolen data reportedly came with each victim’s industry, revenue and headcount attached, like a sales catalogue [6]. In many cases the attackers then reached the company’s central login system — the directory that controls every employee account [6]. As of Wednesday morning, almost all the breached devices were still online [6].

Devices and IP addresses across 194 countries were hit [6]. The angle for anyone running infrastructure: the lesson here is not “patch faster.” It is “rotate any credential that has ever touched the internet, and stop trusting that a default password was changed.” A firewall is supposed to be the wall. When the wall’s own keys are public, everything behind it is, too.

AI splits the rich countries’ club

At the G7 summit in France, the second day belonged to artificial intelligence — and to a new kind of worry [8]. French President Emmanuel Macron told the assembled leaders and AI bosses, including the heads of OpenAI and Anthropic, that the United States could “from one day to the next turn off the switch” on the AI models the rest of the world now relies on [20].

He has a fresh reason. A few days ago the US government ordered Anthropic — an AI company now valued near a trillion dollars — to suspend exports of its two most powerful models on national-security grounds [58]. Other countries watched a single government reach in and switch off a tool their companies had built around [20]. The phrase circulating at the summit was blunt: world leaders want American AI; they just don’t want America to be able to turn it off [20].

Why this matters now: the world standardised on a handful of American AI models the way it once standardised on Windows or the dollar. That convenience is now visible as a dependency — and dependencies can be revoked.

Apple says the price is going up

Outgoing Apple chief executive Tim Cook told the Wall Street Journal that price rises on Macs, iPhones and iPads are now “unavoidable” [50]. The cause is a memory-chip shortage some have nicknamed RAMageddon [27].

The mechanism is a chain. AI data centres need enormous amounts of memory chips, so they are buying up the supply. That competition has pushed memory and storage costs up roughly fourfold in a year, Cook said — a level he called “unsustainable” [27][50]. Apple has been absorbing the cost to shield customers; it has run out of room to keep doing so [50]. He did not say when or by how much prices would rise [50]. The number to watch: memory pricing, because it now ties the cost of a laptop to the building of a data centre on the other side of the world.

The open web tries the oldest trick: email

A quieter story points at a problem the big platforms rarely have to think about. Mastodon — the open, decentralised alternative to X and Threads — shipped version 4.6, and with it, email newsletters [33]. A writer on Mastodon can now send posts straight to subscribers’ inboxes, even subscribers who have no Mastodon account [33].

It is an admission of where the hard part is. Mastodon has about 735,000 monthly active users; the wider open social web it belongs to has over a million [33]. The big platforms have hundreds of millions. By leaning on email — a system everyone already has — Mastodon is trying to reach people without first asking them to join a network almost nobody is on yet [33]. The catch: it costs servers real money to send those emails, so it is off by default [33].

Elsewhere, Pinterest launched an experimental AI shopping app called Ask Pinterest [5], and Anthropic overhauled its design tool to plug into the systems large companies already use — a quiet bet that the way to win is to live inside other people’s workflows rather than replace them [3].

02 · Lesson · why it matters

Why the better tool keeps losing to the one everyone already uses

A thing's worth often comes from who else is already on it — which is why good products fail, switches feel impossible, and one winner takes almost everything.

A network nobody is on

Today Mastodon, the open alternative to X and Threads, added a feature it should never have needed: email. A writer there can now send posts to your inbox without you ever joining Mastodon.

Read that again. A social network is reaching past itself, because the hard part was never the software. Mastodon’s apps are fine. Many people think they are better than the big platforms — no ads selling your attention, no single owner, your account portable to another server whenever you like. And it has about 735,000 monthly users, against hundreds of millions on the networks it competes with.

The product is good. The product is losing anyway. To understand why, you have to look not at the product but at the crowd already standing somewhere else.

The value is the other people

A telephone with no one to call is a paperweight. The second telephone made the first one useful. The millionth made both indispensable. The phone’s worth had almost nothing to do with the phone and almost everything to do with how many other people held one.

This is a network effect, and it is one of the quiet engines of how the world organises itself. The value of a thing rises with the number of people already using it. A language. A currency. A social app. A file format. You don’t join the biggest one because it’s best. You join it because that’s where everyone is — and by joining, you make it a little bigger, and a little harder for the next person to choose anything else.

Mastodon’s problem is now plain. A new social network is only worth your time if your friends are there. Your friends won’t come until it’s worth their time. Everyone waits for everyone else, and the network that already has the crowd keeps the crowd.

Why the worse thing wins

This is the part that should unsettle you, because it cuts against a comfortable belief — that the best tool wins. Often it doesn’t. The tool with the most people already using it wins, and “most people” and “best” are not the same thing.

The keyboard under your fingers was arranged in the 1870s, partly to slow typists down so mechanical typewriters wouldn’t jam. The jam is long gone; the layout remains, because everyone learned it, so everyone teaches it, so everyone learns it. Better layouts exist. They lost. Not on merit — on who got there first and gathered the crowd.

Once a network is large enough, its size becomes the argument. Newcomers don’t weigh features. They count heads. And the count points one way.

The crowd tips, and then it’s over

Network effects don’t just favour the leader. Past a certain point they hand the leader almost everything. This is called tipping: the market slides toward a single winner, because each person who joins makes joining more attractive for the next, until choosing anyone else feels eccentric.

You can see it happening above the apps, too. At the G7 this week, world leaders admitted their countries now depend on a handful of American AI models — and worried aloud that the US could switch that access off. Nobody held a vote to make the world run on the same few systems. It happened the way these things happen: each government and company picked what everyone else was already building on, and the picking compounded into a dependency. Convenience, repeated by millions, hardens into a structure nobody chose and nobody alone can leave.

What this costs the people inside it

Here is the half that’s easy to miss. A network effect is not just a business advantage. It is a wall around everyone standing inside it.

You are inside several right now. The messaging app you use is the one your family uses; you stay not because it’s best but because leaving means leaving them. The document format your work runs on locks you to the software that reads it. These aren’t free choices made fresh each day. They are the accumulated weight of everyone else’s earlier choices, pressing on yours.

That weight is why a writer on a better, freer network is reduced to emailing you directly — quietly admitting the open door can’t compete with the crowded room. It’s why a country that built its economy on someone else’s AI can’t simply walk away. The cost of the network isn’t paid by the loser who built the better thing. It’s paid by all of us, in the doors that quietly close while we’re busy using the one that’s open.

So when you reach for the obvious tool because everyone’s on it, you’re not wrong to. You’re just adding your weight to a wall — one that will, in some small way, stand against the next person who hoped for something better. Knowing that doesn’t free you from the crowd. But it might make you hold your certainty about “the best” a little more loosely. The best thing and the winning thing are rarely the same, and the gap between them is full of doors you never saw close.

03 · Lab · your turn

The Switch

Rehearse why a better network can't pull people away until the crowd suddenly tips.

04 · Hope · carry this

Walls of habit look permanent right up until the moment they tip, and they have tipped before — the crowd that feels immovable today was itself the upstart once. The fact that a smaller, freer network keeps building anyway, reaching for people one inbox at a time, is its own quiet proof that better is still worth making.