Summary of Key Points
Linus Torvalds, the founder of Linux, recently made a clear statement in a community email: The Linux kernel is not against AI. He opposes developers of AI tools; they can either copy the project and fork it themselves or leave. The controversy stemmed from the distribution of responsibility for using an AI code review tool called Sashiko. Linus believes that AI is a useful engineering tool, but the ultimate responsibility must lie with humans. His attitude has shifted from saying in 2024 that "90% of AI was hype" to now acknowledging that "the usefulness of AI is beyond doubt." He also pointed out the burden that AI places on maintainers, with the community divided on this issue.
1. Linus' Firm Stance: Linux Is Not Against AI; Disagreeents Can Choose
Linus was very direct in his message: Linux is not an anti-AI project, and those who disagree have two options—either copy the Linux code and work on their own (a process known as "forking") or simply leave. The controversy began with the use of the AI review tool Sashiko. Some developers argued that human verification was necessary before accepting AI-generated results, while others believed this would diminish the usefulness of AI in helping with maintenance tasks, leading Linus to make a final decision.
He emphasized his role as the chief maintainer: AI, like compilers and code search tools, is just a tool. Although AI is not perfect, there's no longer any debate about its usefulness. "Forking" is a common practice in the open-source community—someone who disagrees with the project direction can simply copy the code and modify it, similar to how Ubuntu originated from a fork of Debian. Linus's message is clear: If you don't want to work with AI, that's fine, but don't get in the way of others.
2. Linus' View on AI: Tools Are Tools, But Responsibility Remains With Humans
While Linus supports AI, he doesn't give it a green light without conditions. His core point is that while AI can be helpful, humans must ultimately take responsibility for its use. For example, the Linux kernel requires that AI-generated code cannot be marked as "Signed-off-by" automatically, as this label represents a developer's commitment to the legality of the code, and it must be signed by a human. When using AI to assist in development, it is essential to clearly indicate which model was used and how. Submitters must also review the AI-generated code to ensure there are no issues.
In other words, AI can be a useful assistant, but it cannot take on the blame for mistakes. Linus also joked about those who criticize AI for making errors, suggesting that "even natural intelligence (humans) isn't always reliable," implying that humans make bugs as well, and we shouldn't deny the value of AI just because it makes mistakes occasionally.
3. A Major Change in Attitude: From "90% Hype to Undoubted Usefulness"
In 2024, Linus claimed that 90% of AI was marketing hype and only 10% actually useful, comparing AI to cryptocurrencies, suggesting that it would take five years to see its true value. Less than two years later, he changed his tune, saying, "There's no longer any debate about the usefulness of AI; those who doubt it probably haven't really used it."
The reason for this change is that AI has begun to show practical results. In 2026, AI was already capable of identifying bugs in the Linux kernel and proposing effective fixes. For example, Greg, a maintainer of the stable version, noted that while earlier AI reports were often inaccurate, now two-thirds of the suggested patches are correct (although they still need to be manually reviewed). Linus dislikes hype, but he acknowledges that AI can indeed solve real problems.
4. The Troubles Caused by AI: Maintenance Workers Under Heavy Pressure
Linus has also observed the negative consequences of AI: The workload for maintainers has increased. For instance, in May this year, the private security email list was filled with bug reports detected by AI, with the same issue being reported multiple times by different AI systems, leaving maintainers overwhelmed. He criticized those who simply pass on raw AI outputs without providing any context or additional information, suggesting that they should first understand the problem and potentially write their own patches before submitting them.
There are also trivial patches generated by AI that may not be necessary for later versions of the software, as making minor adjustments could introduce new issues. Linus argues that not all problems identified by AI are worth fixing, and humans need to decide whether the issue is serious enough and whether it's appropriate to fix it at that time.
5. Divided Community Opinions
Linus' statements have sparked two opposing views within the community:
- Supporters: They see Linus as pragmatic, recognizing that AI technology is irreversible and that it's better to make use of it rather than resist it. For example, they argue that using AI to review code is more efficient than having to manually generate large amounts of code that requires cleaning up.
- Critics: They worry about the potential misuse of AI. Some fear that they might be responsible for fixing bugs generated by AI, and others point out that the term "AI is just a tool" oversimplifies the complexities involved, such as data copyright, energy consumption, and job displacement. Some also believe that media headlines have exaggerated Linus' stance, suggesting that he actually advocates for responsible use of AI.
In summary, while the community acknowledges the usefulness of AI, there is still debate about the responsibilities it entails, the additional work it creates, and the social implications it has.
Final Summary
Linus's position is clear: AI is a useful tool, but humans must control its use. We should not ban it out of fear, nor should we let it offload all responsibility onto us. The path forward for open-source communities with AI involves finding a balance between leveraging its benefits and managing its potential challenges.