Anthropic and OpenAI restrict AI reverse engineering in Claude Code and Codex
Anthropic and OpenAI have introduced security blocks against AI-powered reverse engineering in Claude Code and Codex; OpenAI banned a user who was reverse engineering a Samsung TV, and the Decomp.wtf project shut down. OpenAI says the bans are unintentional and are being fixed.
According to the 404 Media website, Anthropic and OpenAI have introduced security checks that prevent AI-powered reverse engineering in their products Claude Code and Codex. Users report that Claude Code refuses to decompile code and triggers "safety checks", while OpenAI blocked the account of a user who used Codex to reverse engineer a Samsung TV, citing a violation of rules against "cyber abuse". The Decomp.wtf project, which made reverse engineering easier, shut down in the past few days.
The move comes after a wave of popularity for the Reverse Engineer-Anything trend, in which people used AI tools to modify their own electronics — installing new operating systems on smart TVs, unlocking devices bricked by manufacturers (bricked IoT), or removing ads from smart TVs. Tarun Gogineni, an engineer at OpenAI, stated on Twitter that blocking accounts for reverse engineering was not intended and is an unintended consequence of hard-to-set decision boundaries that is being fixed; according to him, there is no ban on reverse engineering in Codex.
Matt Brown, principal consultant at Brown Fine Security, told 404 Media that the blocking of reverse engineering correlates directly with the "cyber" block category and that large AI labs appear to be under pressure to ensure their models are not involved in criminal hacking activity, leading to stricter guardrails. Because of this, Brown is investing in his own hardware and switching to open-weight models — for example, he used the GLM 5.3 model to find hidden Bluetooth commands in a gas sensor connected to a propane tank.
The article's author and the users interviewed expect that AI-powered reverse engineering will not stop, but will shift away from OpenAI's and Anthropic's frontier models toward open-weight, local, and uncensored models, which, in their view, will reduce the availability of this technique for ordinary users.
Why it matters
For individuals who used AI to modify their own IoT devices or smart TVs, this measure means losing a reliable tool and risking the loss of their account without warning. For companies in the security research and pentesting field, it means unpredictable and shifting guardrails on the same model, which complicates long-term investment in these tools and pushes them toward open-weight alternatives, as described by the security consultant interviewed for the article.
Two audiences, two different impacts
What this means
For individuals
Users who used AI to decompile firmware or unlock their own IoT devices and smart TVs are now running into security blocks in Claude Code and account restrictions or bans in Codex, so this approach with large commercial models is no longer reliable.
For a business
Companies relying on Claude Code or Codex for reverse engineering, security research, or pentesting risk a sudden and unpredictable blocking of their account or feature without prior warning, increasing the risk for operational processes dependent on these tools.
Risks and complianceCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.