'Reverse Engineer Anything': AI Companies Crack Down on the Coolest Thing People Were Doing With AI

People were using both Anthropic’s Claude and OpenAI’s Codex to perform AI-powered reverse-engineering, also known as decompiling, which allowed them to unbrick IoT devices, remove surveillance ads from smart TVs, and install open-source or custom software on old hardware. However, several users reported hitting guardrails or having their accounts limited while attempting such tasks.

Several people have mentioned triggering "safety checks" in Claude Code that prevented them from performing AI-powered reverse engineering, resulting in a highly upvoted Reddit post questioning the future of Reverse Engineer Anything. A Discord thread in the Reverse Engineer Anything community reported that Claude refused to decompile anything, with multiple users experiencing similar issues.

OpenAI’s Codex user log4j reported their account was banned for violating Cyber abuse rules while reverse engineering Samsung TVs. The popular project Decomp.wtf, which made reverse engineering software and games easier, had also shut down. Reverse Engineer-Anything enabled people to use AI tools to replace manufacturers' software or operating systems with open-source or custom ones, such as installing new operating systems on smart TVs, turning baby monitors into AI assistants, and unbricking IoT devices abandoned by manufacturers.

Big AI corporations, seeing an opportunity to protect their interests when users circumvent device limitations, immediately took steps to restrict this behavior. As a result, people are moving away from big company models to local, uncensored, and open weight models, as these are less likely to have their accounts disabled or face AI company guardrails.