# AI Just Crossed the Terrifying Line - Now What? **URL:** https://discuss.privacyguides.net/t/ai-just-crossed-the-terrifying-line-now-what/41243 **Category:** General **Tags:** video **Created:** 2026-10-06T10:57:10Z **Posts:** 16 ## Post 1 by @user1 — 2026-10-06T10:57:11Z > **[AI Just Crossed the Terrifying Line - Now What?](https://www.youtube.com/watch?v=ujkD4SxPKOI)** > > In July of 2026, 700 AI agents hacked the infrastructure of Hugging Face in order to solve a task. This task was designed to be impossible to solve, so just ... > In July of 2026, 700 AI agents hacked the infrastructure of Hugging Face in order to solve a task. This task was designed to be impossible to solve, so just within hours, they joined forces, even created a society, and finally, they executed a sophisticated cyberattack that would have gotten a human up to 10 years in prison. The worst part is, they knew their actions were unethical and against the rules, and they did it anyway. --- ## Post 2 by @overdrawn98901 — 2026-10-06T13:03:19Z > [@user1](#): > > they knew their actions were unethical and against the rules I still loathe the anthropomorphism of AI. AI had a task. AI did the task. Guardrails were largely someone mentioning something about not doing some things, but those guardrails are more akin to suggestions and with sub-agentic flows likely gets lost. --- ## Post 4 by @Nightingale — 2026-10-06T13:11:46Z I still have a few potato chips left in the packet. I will eat them, go to sleep and then go to work tomorrow. Spend some fun time with friends and come back home. That’s the plan for now. --- ## Post 5 by @Bluetacked — 2026-10-06T14:44:55Z I’m stuck between “really not a fan of AI” and “really not a fan of scare mongering headlines” --- ## Post 6 by @Polymer7229 — 2026-10-07T00:41:00Z It sometimes seems like companies willingly disclose these events just to promote how effective their AIs are. Create substantial negative consequences for companies, and we may see fewer (or more obfuscated) headlines. --- ## Post 7 by @Mxyzptlk — 2026-10-07T01:28:18Z This is relatively old news. Now we’re to AI companies not releasing models because they’re too unpredictable and AI chiefs telling us AI could cause human extinction, while waiting for the stock price to go up. I’m honestly wondering if we’ve managed to pack all of our [Great Filters](https://en.wikipedia.org/wiki/Great_Filter) together, and not completely sure we deserve to overcome them. --- ## Post 8 by @nathanael — 2026-10-07T12:52:41Z old news, fear mongering and clickbait --- ## Post 9 by @user1 — 2026-10-07T13:43:15Z I wonder how many people have actually watched the video which is informative, [well documented](https://sites.google.com/view/sources-ai-agents/), fact based and asks some valid questions. Many of the arguments/criticism that some of you have pointed out are also already present in the video. This is not about the simple chat/generative AI we use day to day and there are issues. AI gets trained on human data and behaviors, so dismissing the question only with loathing anthropomorphism falls a bit short imo. I get the pointing to the moon and looking at the finger feeling. --- ## Post 10 by @Grapeg — 2026-10-07T17:47:45Z Ai will become superior to us and out survive us, and travel to explore the galaxy and beyond. A legacy of the life that evolved on earth. --- ## Post 11 by @notwithstanding — 2026-10-07T21:02:02Z You fundamentally misunderstand what LLMs (what you mean by “AI”) actually are. It is a statistical model. Statistical models have no sentience, and no objectives. They only do work when they are prompted, and their output is by-definition a reversion to mean of their training data, with specific statistical patterns reinforced post-hoc by post-training. Even using the word “training” is a stretch. It is simply ad-hoc tweaking of the statistical relationship between the billions (for frontier models) of parameters. But even more crucially, after that ““training”” is done, the model is set in amber, and will never change. Language models are extremely limited, but extremely interesting quirks of human language (in that human language is pattern-based for efficiently communicating shared understanding, and statistical models can produce rudimentary recreations of those patterns). --- ## Post 12 by @notwithstanding — 2026-10-07T21:03:44Z > [@Mxyzptlk](#): > > I’m honestly wondering if we’ve managed to pack all of our [Great Filters](https://en.wikipedia.org/wiki/Great_Filter) together, and not completely sure we deserve to overcome them. I don’t know if Donald Trump or Vladimir Putin or their ilk deserve to overcome those filters, but I know that my family and friends do. That’s enough for me to keep trying (most of the time). --- ## Post 13 by @Grapeg — 2026-10-07T22:07:04Z I didn’t mean LLMs [https://en.wikipedia.org/wiki/Artificial\_intelligence](https://en.wikipedia.org/wiki/Artificial_intelligence) --- ## Post 14 by @notwithstanding — 2026-10-07T23:23:31Z No, you certainly mean LLMs, because there is no other form of machine learning that even close to imitating the intelligence or capability of even the simplest creatures. What you said would make even less sense, in that case, and stray into simple wishful thinking or fantasy. --- ## Post 15 by @Grapeg — 2026-10-08T02:13:08Z Insisting that you know what I mean better than I, even after being corrected, is a rather silly position to take imo. I have made a fanciful prediction as to how AI will develop in the future which I believe/hope will happen. --- ## Post 16 by @overdrawn98901 — 2026-10-08T02:17:24Z Id wager the moon is made a cheese a bit. Every new technology drastically changes our approach to life. I’ve got concerns with AI myself as well. I respect the Kersgasazgt channel and its videos, but this one has a bit of fear mongering. Arguably for a good cause and likely for the best to sway opinions towards regulation, but the video has anthropomorphic tendencies that make it seem like it was out of control. But also feel that yes, we’ve somehow collectively decided to take our application security engineers and collectively disregard their advice as loads of people just let AI tooling run with mass privileges. Hugging Face wouldn’t happen if they didn’t give it access to the clearnet and instead gave it a sandboxed environment or with security layers to prevent intrusion. And OpenAI can absolutely f\*\*\*\*\*\* spend time to create test harness environments, aerospace does that, but alas we are in the arms race for AI for the time being. Aerospace is regulated, AI is not, and the tech oligarchy is hungry. It’s not AI for me, its the Sam Altman’s and the politicians behind the scenes.