AI safety checks hit reverse engineering of owned devices as Decomp.wtf shuts down
Practitioners are weighing open-weight and local models as frontier labs' guardrails tighten around reverse-engineering work.
Users of Anthropic's Claude Code and OpenAI's Codex who used AI to reverse-engineer devices they own, such as smart TVs and internet-of-things hardware, have recently hit safety checks, warnings or account restrictions. One OpenAI account was reportedly banned over its rules on cyber abuse, and the Decomp.wtf project that simplified reverse engineering of software and games has shut down. An OpenAI engineer said the blocks are mostly accidental and that there is no prohibition on reverse engineering with Codex, adding that the issue is being fixed. Some practitioners say they are shifting toward open-weight and local models, which they expect to be more stable than frontier labs' guardrails.
Why it matters
For people who legitimately take apart and study devices they own, the tools they relied on have become less predictable. Users of Claude Code and Codex have reported safety checks and warnings, and one account was reportedly banned. An OpenAI engineer has said the blocks are mostly accidental and that the company is fixing them, but the uncertainty has already pushed at least one security consultant toward open-weight models. The episode shows how much of this work now depends on decisions made by the companies that host the models.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.