Users of Anthropic Claude and OpenAI Codex have encountered limitations when attempting to apply AI to reverse engineering. Security mechanisms also began to block software decompilation, and in some cases accounts were blocked. Even projects related to the modification of devices owned by users themselves were subject to restrictions.
According to 404 Mediaseveral Claude Code users reported triggering security mechanisms that prevented software from being reverse engineered, and the Reverse Engineer Anything community on Reddit posted the title “The Dream Is Dead,” stating that it would probably no longer be possible to use Claude for similar tasks. Other members of the Discord community also confirmed that the service refuses to decompile programs.
Codex users encountered similar problems. One reported that OpenAI suspended his account after reverse engineering Samsung TVs, citing violations of cybersecurity abuse rules.
Interest in this area has grown thanks to the free and open source project Reverse Engineer Anything, which uses AI to analyze electronics and replace factory software with open or AI-powered alternatives. Users installed new operating systems on smart TVs, turned them into AI assistants, repaired Internet of Things devices that manufacturers had stopped supporting, and updated the software of older electronics.
404 Media also draws attention to the contradiction in the policies of large AI companies. In particular, technology corporations protect the interests of electronics manufacturers by limiting the use of AI for reverse engineering. At the same time, collecting huge amounts of materials from the Internet to train their own models does not cause large companies to have a similar attitude towards copyright issues.
OpenAI spokesman Tarun Gogineni said that account bans for reverse engineering should not occur. According to him, most of these restrictions arise by chance due to the complexity of setting up AI protective mechanisms, and the Codex itself does not prohibit such activity. Gogineni added that the problem is being resolved.
However, some experts are already choosing to switch to alternative models, fearing sudden changes in the rules. Matt Brown, principal consultant at Brown Fine Security, told 404 Media that reverse engineering blocks are almost exactly the same as triggering cybersecurity filters.
In his opinion, developers, in their desire to prevent the use of AI for hacking, impose overly strict restrictions, which interfere with legitimate research. Brown intends to invest in equipment and local models with open scales, which are becoming smarter, including thanks to the developments of Chinese companies.
If you notice an error, select it with the mouse and press CTRL+ENTER.
