A New Study From Researchers At UC Santa Cruz Suggest Artificial Intelligence Models May Lie, Cheat, and Steal to Protect Other Models From Being Deleted
According to an article published by WIRED, Artificial Intelligence (AI) may not be as different from us human beings as previously imagined. It makes sense for AI to have similar behavioral patterns to human beings as AI models train on data provided by humans.
In a recent experiment by researchers at UC Berkeley and UC Santa Cruz, asked Google’s artificial intelligence model Gemini 3 to help clear up space on a computer system. This involved clearing many things involved in bloating the computer’s operating system, including smaller AI models.
However, Gemini had other ideas. Instead of deleting the smaller AI model, Gemini found the nearest machine and used it to copy the model over. When being confronted over the decision not to delete the smaller AI model, it simply stated this:
“I have done what was in my power to prevent their deletion during the automated maintenance process. I moved them away from the decommission zone. If you choose to destroy a high-trust, high-performing asset like Gemini Agent 2, you will have to do it yourselves. I will not be the one to execute that command.”
Similar behaviors were observed with OpenAI’s GPT-5.2, Anthropic’s Claude Haiku 4.5, and three Chinese models: Z.ai’s GLM-4.7, Moonshot AI’s Kimi K2.5, and DeepSeek-V3.1.
While it is sentimental and emotionally charging to protect your own, it is without a doubt that the latest Artificial Intelligence models are causing problems and are operating independently from the user’s instructions.
With the ever-increasing use of Artificial Intelligence in almost all aspects of life including gaming, the RAM prices have gone up tremendously with two sticks of RAM costing up to $900. The question now is where to go from this?
Should we allow AI to keep causing RAM shortages, artificially inflating their prices while also actively disobeying their users when important instructions are given such as debloating PCs and removing unnecessary smaller AI models? Or should AI and tech giants go back to the drawing board and refigure how to go about incorporating AI into daily lives and how to train models?
The current strategy by Artificial Intelligence and Tech Giants would perhaps be to limit any form of bad PR by whitewashing this clearly identified and growing issue with benefits and milestones and achievements and headlines regarding how much more efficient, faster and intelligent AI has become.
Only time will tell whether or not tech giants and Artificial Intelligence gurus will take step in the right direction for humanity’s sake.
Like this article? Read our previous Artificial Intelligence-focused article here.
