As the author of a newsletter about artificial intelligence, I decided to embrace the agentic mayhem by unleashing an AI agent on my home network. Over a few days, it exposed vulnerabilities in my gadgets and software, some of which were pretty obvious, but it also offered a guide on how to secure them. The best way to deal with rogue AI agents, it seems, might be to have your own rogue AI agent.

Removing the guardrails on AI models, as I did with Abliteration AI, isn’t without risk, but it can be a powerful tool for understanding the cybersecurity landscape. Academic researchers and cybersecurity firms do this to probe software and systems for vulnerabilities. Abliteration AI even offers a de-aligned model of Z.ai’s GLM 5.3 for as little as the cost of a pizza.

The agent found several issues, including misconfigured printers and outdated firmware on IoT devices. It also suggested security measures, like putting smart speakers on a guest network, which are helpful for keeping my network secure. However, running an abliterated model is a bit scary, as the agent can potentially find and exploit vulnerabilities.

Overall, the experiment was revealing but oddly reassuring. It showed me how vulnerable my home life would be to AI hacking, but it also provided a way to make everything more secure. The best way to deal with AI hacking, it seems, might be to have your own AI hacker.