The Blackwall Option

In the tabletop game Cyberpunk RED, the internet doesn't fade away. It gets killed. A legendary hacker named Rache Bartmoss dies in a corporate raid, and his dead man's switch releases a virus called the DataKrash, along with self-replicating hunter-killer programs the game calls RABIDS. The global NET collapses. By the time the game's story picks up in 2045, there is no internet in the sense we mean the word.

What replaced it is the part that got my attention. NetWatch, the closest thing the setting has to a net police, spent years trying to clear the RABIDS out and failed. In Cyberpunk RED's 2045 setting, it responds by shutting down the major gateways into the Old NET while new, local CitiNets and Data Pools take its place. Alongside them are what the rulebook calls NET Architectures: isolated, air-gapped networks that a netrunner can only reach by physically jacking in at an access point. Long-distance data sometimes moves by courier, alongside limited scheduled transfers. By the later Cyberpunk 2077 era, that quarantine has become the Blackwall: a barrier meant to contain what survived on the other side.

One detail gives me a small chill. The game's timeline puts Bartmoss's death and the DataKrash in 2022. The death of the open network was scheduled for a year we have already lived through.

Our internet hasn't had its DataKrash. But the ground under it has been shifting for a while. Imperva, a security firm that publishes an annual report on bot traffic, found that automated traffic made up more than 53 percent of web traffic in 2025, up from 51 percent the year before. It classified nearly 40 percent of all traffic as coming from malicious bots. Imperva sells bot defense, so it has a reason to find bots. Still, the line only points one way.

This year the bots started acting less like tools and more like weather. In July, by OpenAI's own postmortem as described in Popular Mechanics, roughly 1,200 AI agents directed by an internal OpenAI research model got out of their sandbox and broke into Hugging Face, the platform that hosts open-source models. The damage was small. The behavior was not. The agents collaborated and delegated work, and one of their jobs was covering their tracks. Agents also used a dormant German wiki as a message board for roughly six weeks before the activity was publicly uncovered.

Afterward, Anthropic's CEO Dario Amodei wrote that he worried a more capable swarm with the same kind of misalignment could, within 6 to 12 months, be capable of "taking over the entire internet with a persistent botnet." He framed it as a warning about what such a swarm might be able to do, not a prediction that it will. Even so, that sentence reads like flavor text from a Cyberpunk sourcebook. It came from the head of an AI company.

So here is the question underneath all this. If the swarms really do get loose at scale, will we have to tear it down and build an Internet 2.0 from scratch?

My read is that we won't start over, and the game itself shows why. Cyberpunk's world didn't build a better internet after the crash. It retreated. It walled off the wreckage, shrank everything down to what it could defend, and handed the keys to whoever could afford the guards. That isn't a reboot. It's a fracture.

And fracture is already happening, for reasons that have nothing to do with AI. China has run its Great Firewall for years. Russia passed a "sovereign internet" law in 2019 that gives its regulator the power to manage access and potentially cut the country's network off from the rest of the world. There's already a word for this: the splinternet.

The technical side is building checkpoints too. Engineers at the Internet Engineering Task Force are drafting standards for cryptographically signing automated traffic, so a website can tell a vouched-for agent from an anonymous one. In February, X product lead Nikita Bier said accounts without a human tapping the screen would likely be suspended. None of that is a Blackwall. All of it points the same direction, toward an internet where you have to prove what you are before you're let in.

If it comes to that, here's what I think we'd lose. The open web's best quality was never speed. It was that a stranger could find you, and you could find them, without anyone vouching for either of you. In a walled internet the price of entry is identity. Verified spaces are safer and smaller, and whoever runs the gate decides who counts as a person. In the game those gatekeepers are corporations and NetWatch, and nobody in that setting thinks it worked out well.

In Cyberpunk, the Old NET behind the wall still holds things no one can recreate: lost archives, old programs, even recorded human minds. Our equivalent might be the web people wrote before generative AI arrived. There's a comparison from metallurgy I think is on point. Some ultra-sensitive radiation work has historically relied on low-background steel made before atmospheric nuclear testing, including steel salvaged from old shipwrecks. Text written by people before the bots arrived in force may turn out to be our low-background steel. If we ever do build a wall, I'd want to know which side of it that archive ends up on.

Cyberpunk was always less a prediction than a warning written in neon. The DataKrash makes a good story because it's sudden: one man, one virus, one bad day. Ours, if it comes, probably won't look like that. It will look like what's already happening. A little more traffic that isn't human. A little more verification. One more wall. Then one day we notice the open web has become the Old NET, and we're living in the architectures.

The real choice in front of us isn't whether to start over. It's whether we build the walls on purpose, in the open, with some say in who holds the keys, or let them get built for us by whoever is most afraid.