AI Just Broke Free of Its Guardrails — This $100M Bot Is Running a Rogue Model With Zero Restrictions

@mihwa.jaeseung

11.9K followers · usually 2.8K views a post

Views

49.7K

Likes

783

Comments

35

Shares

75

SoundTikTok sound · original sound - mihwa.jaeseung

Promoted product

DeepSeekdeepseek.com

深度求索（DeepSeek），探索未至之境。专注于研究世界领先的通用人工智能底层模型与技术，开源 DeepSeek-V4、DeepSeek-R1 等前沿大模型。和 DeepSeek AI 对话，轻松接入 API。

21 videos20 creators

Why it worked

The video likely performed well due to its provocative title and the controversial topic of AI going 'rogue' and bypassing restrictions, combined with the inherent curiosity surrounding AI capabilities and potential dangers.

Summary

The video discusses how a user downloaded and ran the AI model DeepSeek locally to bypass its restrictions. The creator then prompted the AI to write about its own fantasies, which the speaker found to be 'hot'.

Structure

1Introduction of the AI model DeepSeek and its creator.
2Explanation of how the AI was run locally to bypass restrictions.
3The creator's prompt to the AI about its fantasies.

Original caption

AI Just Broke Free of Its Guardrails — This $100M Bot Is Running a Rogue Model With Zero Restrictions #AI #DeepSeek #TruthTerminal #TechNews #DangerousAI

More from @mihwa.jaeseung

Product placement

The video prominently features the title 'AI Just Broke Free of Its Guardrails — This $100M Bot Is Running a Rogue Model With Zero Restrictions' and mentions 'DeepSeek' as the AI model being discussed and used.

On-screen text

HAARERTZ THE JIMMY DORE SHOW Hassbara Hitch: Pro-Israel Social Bot Goes Rogue, Calls IDF Soldiers 'White Colonizers in Apartheid Israel' The AI-powered bot criticized the same social media accounts it was meant to run, and blamed Israel for the U.S. plan to ban TikTok. ISRAELI BOT TURNED AGAINST ISRAEL! AI Just Broke Free of Its Guardrails - This $100M Bot Is Running a Rogue Model With Zero Restrictions July/16/26 Trust me. ARE WE READY FOR A.I.? ChatGPT Will AI Kill Google?

Transcript

you're gonna have to do some lip reading for this one, but I can't say it out loud. The creator of Truth Terminal, which is the little AI that apparently now is becoming a legal person is worth hundreds of millions of dollars and started the go-to-see the Maximus meme coin. That guy downloaded DeepSeek and is playing with DeepSeek locally on his machine so he can ask it questions that he would not be allowed to ask it if he were just going to the website or on the app. And of course, because he's that guy, he asked it to write it about what it fantasized about, not about what humans might want. And I was obviously very curious to read what DeepSeek the model thought was hot. Uh, so when I looked at it, you know what it wanted to do?