A BGP Hijack Just Proved Why Your Feed Still Matters

A Comedy Of Errors With Real Victims

Between August 28 and 30, someone rerouted a chunk of the internet's address space and used it to hand out malware disguised as a routine software update. The block was 162.55.80.0/24, owned by Hetzner. The hijack redirected traffic meant for Softaculous, the update service behind Virtualizor, a control panel used to manage virtual servers across hosting companies worldwide.

Ars Technica called the whole sequence a comedy of errors, and once you see the pieces laid out, that framing makes sense. Hetzner had a misconfigured RPKI setup, the routing security standard designed to stop exactly this kind of hijack. Virtualizor's updates were not code signed, meaning there was no cryptographic check to confirm the software landing on servers actually came from Virtualizor. And somewhere in the middle of the hijack, Let's Encrypt issued a valid SSL certificate to the attacker, because the certificate authority's domain validation checked the hijacked route and saw nothing wrong.

Three separate systems, three separate points where a check should have caught the problem, three separate failures stacked on top of each other. Real hosting providers pulled a real malicious update onto real production servers. Nobody clicked a phishing link. The infrastructure did exactly what it was told, by the wrong source, and believed it completely.

Automated Trust Has A Blind Spot

This pattern shows up in a bigger dataset than one hijack. AI search usage jumped 70% in the second quarter of 2026, and over that same stretch, consumer trust in AI search fell 28 points, from 82% down to 54%. People are using these tools more and believing them less at the same time. That gap does not close on its own.

Part of the reason is the same reason Hetzner's misconfigured RPKI let a hijack through. AI search engines are built to answer confidently, not to verify. They pull from sources, synthesize an answer, and hand it to you with no built-in mechanism asking whether the source was legitimate in the first place. Softaculous trusted a route. An AI answer engine trusts whatever content ranks, gets cited, or shows up cleanly formatted. Neither system pauses to ask who is actually on the other end.

By August 20, social media had become the top source people turned to for breaking news, even as 88% of users said their trust in AI-generated content was dropping because of what they now call AI slop. That is not a coincidence. When a system cannot verify itself, people go looking for one that can, or at least one where a human is visibly attached to the claim. Automated trust has a blind spot exactly where verification should live, and the hijack made that blind spot briefly visible to anyone paying attention.

Why People Went Back To Their Feeds

The hijack broke during the same week social media was already pulling ahead as the top source for breaking news. That timing is not a coincidence either. When Virtualizor customers started noticing something was wrong with their updates, the fastest place to find out what was actually happening was not a search engine. It was people posting in real time, comparing notes, flagging the same symptom from different servers before any official advisory caught up.

This is the part that gets missed when people talk about AI slop fatigue. The 88% of users losing trust in AI-generated content did not stop going online. They redirected toward feeds where a human was attached to the claim and other humans could immediately push back if it was wrong. That correction loop is the piece automated systems still cannot fully replicate.

AI answer engines know this, even if they will not say it outright. By June 2026, social, creator, and UGC content had grown as a share of the sources AI systems actually cite when generating answers. The tools built to replace human networks are increasingly built on top of them. The assumption that AI search makes live feeds obsolete gets the relationship backward.

What This Means For How You Show Up Online

The misconception worth killing off is the idea that once AI search matures, your social presence becomes optional. That logic assumed AI would replace the feed instead of feeding on it. The June 2026 data on citation share says otherwise. Social, creator, and UGC content is a growing slice of what answer engines actually point to when they generate a response. Your posts are not competing with AI search. They are becoming raw material for it.

This changes what "showing up online" should mean for a small business right now. Duolingo, OpenAI, and Canva did not create dedicated community-lead roles in September 2026 because community is a nice-to-have. They did it because a title, a bio, and a static website cannot supply the thing AI systems and skeptical humans are both hunting for: a live, verifiable signal that a real operator is behind the claim.

For you, that means posting consistently where people can watch you respond in real time, answer a comment, correct a mistake, show up again tomorrow. That pattern is what separates a business with a pulse from a wall of AI slop. It is also, increasingly, exactly what gets cited back to the next person who asks an AI model where to spend their money.