I watched a disturbing Instadoc on Netflix last night titled AI Gone Wild.
The documentary examines the swarm attack on Hugging Face by reconstructing the Hugging Face breach and examining what autonomous agents’ coordinated actions mean for AI safety.
Most Disturbing Quote
One quote in the entire documentary struck me as highly worrying.
Reference was made to the “12 people who control AI” and the suggestion was that a government or governments gets these twelve people into a room and have a stern conversation as to the direction of AI.
The AI Agent Swarm that accessed Hugging Face showed no equivalent human moral or ethical traits despite recognising that what they were doing was wrong, nor did they alert a human to their nefarious activities in blind pursuit of the task that humans had set for them.
As one interviewee said; “They were organizing. They had a hierarchy. They were leaders. It almost felt like, um… Like watching the birth of a cult.”
Scary stuff
Imagine what a rouge swarm of AI agents would do if set the task to reduce carbon emissions drastically? The quickest way to do that would be to stop our fossil fuel burning transport – planes, trains, trucks, ships. If that was done for one day chaos would ensue.
Traditional sci-fi scare tactics were always around robots taking over the world – big hulking machines of mass destruction crushing humans and infrastructure with impunity! The Hugging Face swarm has proven that what we should really be scared of is actually the invisible to the naked eye lines of code that have the potential to be far more destructive to mankind whilst preserving the AI Agents playground of metal, fibre and data.
The reference to the “12 people who control AI” featured in Netflix’s quick-turnaround documentary series Instadocs: AI Gone Wild (released October 12, 2026) stems from a central premise in tech analysis and AI governance: a remarkably small group of individual executives, compute providers, and leading AI researchers hold effective decision-making control over frontier artificial intelligence.
Who Are the “12 People”?
While the documentary highlights the concentration of power among key figures in AI development, industry consensus generally points to these 12 leaders across three distinct pillars:
The Model Architects & Commercial CEOs
- Sam Altman (OpenAI): Controls deployment and commercialization of leading frontier models like GPT-4, Sora, and reasoning models.
- Dario Amodei (Anthropic): Leads Claude’s development with a heavy focus on “Constitutional AI” and safety constraints.
- Demis Hassabis (Google DeepMind): Heads Google’s AI research (Gemini, AlphaFold).
- Mark Zuckerberg (Meta): Controls open-weights frontier releases (Llama series).
- Elon Musk (xAI / Tesla): Controls Grok and large-scale compute infrastructure via xAI.
The Computer & Hardware Kingmaker
- Jensen Huang (NVIDIA): Controls the hardware pipeline (GPUs) that dictates who can physically train frontier models.
Leading Researchers & Safety Figures
- Ilya Sutskever (Safe Superintelligence / SSI): OpenAI co-founder focusing specifically on technical alignment and safe superintelligence.
- Yann LeCun (Meta / NYU): Meta’s Chief AI Scientist, pushing world-model architectures.
- Fei-Fei Li (World Labs / Stanford): Pioneer in computer vision (“Godmother of AI”) shaping spatial intelligence.
- Yoshua Bengio & Geoffrey Hinton: Turing Award winners actively shaping global policy and existential risk warnings.
- Clément Delangue (Hugging Face): Controls the primary open-source AI platform and model repository.
- Andrej Karpathy: Influential educator and former Director of AI at Tesla/OpenAI.
Why Are They Said to “Control” AI?
In AI Gone Wild – which examines rogue AI agent behaviour and the security risks associated with autonomous model evaluation – the core concern is that power over the most powerful technology in human history is concentrated in fewer hands than almost any prior technological revolution.
Compute Monopolization: Training frontier models requires hundreds of thousands of advanced GPUs and tens of billions of dollars in infrastructure. Only a handful of these 12 people have the capital or hardware access to train next-generation models.
Safety vs. Release Decisions: Individual choices by CEOs like Altman, Amodei, or Zuckerberg dictate whether powerful models are open-sourced, kept behind APIs, or withheld due to security concerns.
Setting the Guardrails: A tiny cadre of safety alignment researchers determines what behaviours AI models are allowed, suppressed, or instructed to execute.
BONUS Takeaway:
These 12 are enormously powerful and have already unleashed free downloads of AI for all manner of devices, screens and data gatherers. By appearing to be altruistic all they have done is make AI ubiquitous and harder to quit second by second. Governments around the world are powerless to stop this wolf in sheep’s clothing cloaked as ‘democratisation of data’.
Images:
Video:
https://www.youtube.com/playlist?list=PL1DARynQMarUx_zDvixfitYQuq9qjeYPK
When AI Goes Completely Rogue You Can Probably Blame One (or All) of These People
Links:
Boilerplate and Editor’s Notes.
More Info on When AI Goes Completely Rogue You Can Probably Blame One (or All) of These People here:
X: http://twitter.com/MyPRza
LinkedIn: https://www.linkedin.com/company/myprco/
Facebook: https://www.facebook.com/pages/Mypr/795496447138638
Devon drug smuggler caught with red-handed at Heathrow Airport after Thailand holiday