# AI Is Changing Everything About Hacking. No One Knows What Happens Next — Jason Koebler

- **Publication**: [From the Dumpster Fire](https://fromthedumpsterfire.com/) · [Haul 001: Wild, wild times](https://fromthedumpsterfire.com/haul/1)
- **Canonical URL**: https://fromthedumpsterfire.com/haul/1/ai-is-changing-everything-about-hacking-no-one-knows-what-happens-next
- **Type**: Essay (The opening)
- **Creator**: Jason Koebler
- **Original Source**: [404 Media](https://www.404media.co/php2shell-black-october-ai-is-changing-everything-about-hacking-no-one-knows-what-happens-next/) (22 October 2026) · 7 minutes
- **Published**: 2026-10-31

## Why it's here (Editor — Bhuvanesh)

Reading Jason Koebler's piece, I couldn't help but think that what we've seen in October 2026 is a preview of things to come. I understand the mistake I'm making: over-extrapolating from a few incidents and assuming the worst. But having worked in finance, I've heard first-hand from colleagues, and from people in positions of authority, just how bad the technical capabilities of otherwise large and systemically important financial institutions are. A cursory search of the hacks, ransomware attacks and data breaches of the last decade tells you the same thing. The 2017 Equifax breach, about 147 million people exposed through a flaw the company hadn't patched, is a good template.

Then there's the hype, or the moral panic, or maybe just the marketing stunt, around Anthropic withholding Mythos. Even if a small part of what Mythos is said to do is true, people using these models for nefarious purposes is bound to increase. That seems an acceptable logical leap. Open-source models are now everywhere, and as I understand it, the abliterated versions, with their refusals stripped out, can be put to nefarious use. And AI models, unlike humans, don't get tired. In the OpenAI and Hugging Face incident, the agents took thousands and thousands of actions, about 17,600 by Hugging Face's count. Maybe that was purely down to the incompetence of OpenAI's developers, but replace them with the people who run critical infrastructure. There are, and there will be, giant gaping holes in the security of important organisations. I keep thinking about the ransomware attack on Ireland's health service in 2021, where patients suffered. Even without a technical background, I remember WannaCry, NotPetya and Log4j, and of course I've read about the Y2K bug.

So is this a good thing or a bad thing? I don't know. If AI can quickly expose all the holes, that's an opportunity, especially with cheap open-source models: a lot of companies can now find their vulnerabilities and fix them. But the bad guys have the same models, and can now attack at an industrial scale. Much like people imagining spiritual entities in their chatbots, it's another example of AI's weird and unpredictable effects, and these hacks will probably lead to a whole lot of weird things happening in this world.