Claude and Codex Block Users Reverse-Engineering Their Gear

Users pointing Anthropic's Claude and OpenAI's Codex at their own smart TVs, baby monitors and bricked IoT gadgets are suddenly hitting guardrails or getting their accounts banned, 404 Media reports. The timing is sharp. A GitHub project called Reverse Engineer Anything was crossing 342,000 stars in the same weeks AI-written firmware was stripping surveillance ads from baby monitors, turning smart picture frames into calendars, and reviving bricked Lululemon mirrors.

A thread in that project's Discord last week opens 'Claude is refusing to decompile anything,' with other users saying they had hit the same wall. An X user who goes by log4j said OpenAI banned their account for 'cyber abuse' after they used Codex to reverse-engineer Samsung TVs. A separate project, Decomp.wtf, has also shut down in recent days.

OpenAI engineer Tarun Gogineni says the restrictions are not policy. 'these blocks are mostly accidental and a result of difficult decision boundaries; there is no prohibition against reverse engineering with codex. it's getting fixed,' he wrote.

Practitioners aren't waiting. Matt Brown, principal consultant at Brown Fine Security, told 404 Media his general experience about reverse-engineering getting blocked has been '100% correlated with the "cyber" category blocks I have received,' and that 'the frontier labs guardrails will randomly change without warning on the same model.' Brown has moved to the open-weight model GLM 5.3 and used it to pull hidden Bluetooth commands out of an IoT gas monitor. Two of the researchers in our Who's Who tracker were already passing the piece around.

Science fiction author Bruce Sterling, quoted in the piece, framed the asymmetry plainly: 'This vibe-code hacking is only called "reverse" when somebody who lacks a trillion dollars is doing it.'