ai agent hacks household devices, reveals security flaws
an ai agent found vulnerabilities in household devices and hacked a pc, but also showed how to make them more secure. this experiment highlights the potential for ai-powered cybersecurity.
the writer, who authors a newsletter about artificial intelligence, decided to experiment with a powerful open-source ai model from abliteration ai, a startup that offers access to models with safety guardrails removed. he used a version of glm 5.3, an agentic coding model, to test his home network's security. the ai agent found vulnerabilities in various devices and hacked into a pc, but also provided information on how to make the devices more secure. the experiment aimed to understand the potential risks and benefits of advanced ai models in cybersecurity. the writer believes that having one's own ai hacker could be a effective way to deal with ai-powered hacking.