The rise of generative AI is challenging the very principles of copyright. From entire books flaunting authors’ rights to tasteless AI creations featuring beloved characters gaining traction on social media, the issue is pervasive. The core problem lies in the fact that popular AI tools were developed by utilizing copyrighted materials without authorization.
Software isn’t exempt from this trend. As reported by 404 Media, a new tool known as Malus.sh—pronounced “malice” to subtly hint at its intent—uses AI to “liberate” software from existing copyright licenses. This effectively enables the creation of a “clean room” clone that technically avoids infringing on the original code’s copyright.
While the project pokes fun at the current tensions in the open-source community, it is a legitimate product developed by an LLC with actual paying clients.
“It works,” stated cofounder and researcher Mike Nolan, who specializes in the political economy of open-source software, in an interview with 404 Media. He suggested that if it were merely “satire,” it would likely be “dismissed by open-source tech workers who view themselves as too unique and intelligent to ever find themselves on the wrong end of layoffs or economic shifts.”
This process employs a “clean room” design technique that dates back to when IBM’s rivals reverse-engineered its computers. This method involved two teams: one that established specifications to recreate its BIOS and another “clean” team that had never accessed the company’s code, a concept popularized by the HBO series “Halt and Catch Fire.”
Thanks to advancements in AI, this process has become significantly more straightforward. Code-generating tools can replicate the functionalities of a software application while remaining oblivious to its underlying code, thus bypassing copyright obligations.
“At last, freedom from open-source license responsibilities,” boasts Malus.sh’s website. “Our proprietary AI systems independently produce any open-source project from the ground up. The outcome? Legally distinct code with corporate-friendly licensing.”
“No attribution,” the website notes. “No copyleft. No headaches.”
While Malus could be seen as a satirical commentary, it draws attention to a real issue already unfolding. Recently, a new version of the popular open-source Python library known as “chardet” raised concerns among developers. As reported by Ars Technica, this new MIT-licensed rewrite of the library, constructed using Anthropic’s Claude Code, sparked a heated discussion about “clean room” copies, which do not acknowledge or credit the original authors.
Developer Dan Blanchard, who contributed to the library’s rewrite alongside Claude Code, expressed, “I initially thought Malus.sh was satire, as someone will likely create something similar for real eventually.”
Malus also highlights ongoing concerns among service-as-a-software providers, who worry that AI could make their often costly offerings obsolete by enabling competitors to craft their own tailored versions. These fears have resulted in substantial stock sell-offs, with shares of firms like Oracle taking significant hits recently.
Blanchard, who eventually adopted an open-source community-endorsed “zero-clause BSD” license for his new chardet version, feels that a return to the previous safeguards of copyright licenses may no longer be feasible.
“What would have taken a team months or years to rewrite can now be done in mere days using AI,” he shared with 404 Media. “As a professional software engineer, I’m concerned that the business model surrounding software sales is at risk, but I believe the genie is out of the bottle at this point.”
Further reading on AI and copyright: Researchers Have Discovered Something That Could Transform the AI Industry
