AIBRIEF

Linus Torvalds Rebukes Anti-AI Stances in Linux Kernel Code Review, Says 'Linux Is Not One of Those Anti-AI Projects'

E
EffectStory 編輯部Editorial Team
Published · Updated
According to Tom's Hardware, Linus Torvalds told the Linux kernel mailing list that Linux 'is not one of those anti-AI projects' and that dissenters can fork the project. According to ITHome, kernel rules still require human reviewers to verify AI-generated code and take full responsibility for it.

What is Linus Torvalds's position on the Linux kernel adopting AI tools?

According to ITHome, Torvalds posted on the linux-media mailing list to make clear that the Linux kernel will not become a project that rejects AI tools outright (E1). He said AI is a development tool like any other, and that developers can decide individually whether to use it — but the kernel community will not ban large language models simply because some contributors object (E1).

In a separate, longer message on the Linux kernel mailing list reported by Tom's Hardware, Torvalds went further: "I realize that some people really dislike AI, but this is an area where I'm willing to absolutely put my foot down [...] Linux is not one of those anti-AI projects, and if somebody has issues with that, they can do the open-source thing and fork it. Or just walk away" (E9). He added that "AI is a tool, just like other tools we use. And it's clearly a useful one. It may not have been that 'clearly' even just a year ago, but it's no longer in question today" (E13) — a shift from his 2024 dismissal of AI tools as overhyped, as noted by Tom's Hardware (E13). Torvalds also said the tool in question "keeps finding embarrassing bugs," and that he "will very loudly ignore people who try to argue against other people from using it" (E14).

What do current Linux kernel rules require for AI-assisted development?

Despite Torvalds's openness to AI, ITHome reports that existing kernel norms still place responsibility on humans. Current rules require that human submitters review AI-generated code, confirm licensing, and take full responsibility for it (E2). The kernel's AI-assisted development documentation specifically states that AI agents cannot add a Signed-off-by tag, because only humans can provide the legal certification required under the Developer Certificate of Origin (DCO) (E7). Instead, the documentation requires an Assisted-by tag whenever an AI tool helps with development, identifying the tool or agent name and model version used (E8).

How does Sashiko, the AI agentic review system, perform?

The debate was triggered by discussion over how to use Sashiko, an open-source AI agentic code review system built by the Linux Foundation, according to ITHome (E3). Sashiko monitors the Linux kernel mailing list for new patches and checks them through a multi-stage review process covering architecture, security, resource management, and concurrency issues (E3).

According to Tom's Hardware, Sashiko's project page claims the tool can find 53.6% of bugs in proposed patches, and argues that this metric already places it above human-level review, since the patches in question had already gone through human review (E10). The tool's false-positive rate is harder to measure and is pinned at within 20% (E11).

MetricValueSource
Bug detection rate on proposed patches53.6%Sashiko project page, via Tom's Hardware (E10)
False-positive rateWithin 20%Sashiko project page, via Tom's Hardware (E11)
Torvalds's prior stance on AI tools2024 (called overhyped)Tom's Hardware (E13)

What disagreement exists among Linux kernel developers over how to use Sashiko?

According to ITHome, kernel developer Laurent Pinchart argued that maintainers who plan to act on Sashiko's review findings should first classify and verify the results themselves, rather than forwarding unconfirmed AI reports directly to contributors (E4). Sashiko developer Roman Gushchin countered that requiring maintainers to first classify and verify every finding would undermine the tool's purpose of helping maintainers (E6).

Tom's Hardware reports that Gushchin, identified as a Google engineer and one of Sashiko's creators, pointed out that Pinchart's proposed approach would undermine the tool's utility, and characterized Pinchart's position as "quite anti-LLM" — a characterization Torvalds echoed in his reply (E12).

What position have developers opposed to AI tools in the Linux kernel taken?

According to ITHome, Pinchart cited generative AI usage guidance published by the Software Freedom Conservancy, which argues that free and open-source software communities should support contributors who choose not to use AI systems, and that developers should not be forced to adopt such tools (E5).

Tom's Hardware notes that AI-generated "slop" code has been a problem for some other open-source projects, specifically naming Gentoo Linux, Curl, and Ghostty, which have limited or outright banned LLM-created contributions (E15).

What does this mean

Torvalds's statement that Linux "is not one of those anti-AI projects" (E9) and his instruction to "loudly ignore" people arguing against AI tool use (E14) puts him on record against the position Pinchart voiced by citing the Software Freedom Conservancy's guidance to support AI-refusing contributors (E5). Gushchin's characterization of Pinchart's stance as "quite anti-LLM," echoed by Torvalds (E12), sits alongside the fact that other open-source projects — Gentoo Linux, Curl, and Ghostty — have moved in the opposite direction by limiting or banning LLM contributions altogether (E15). Yet even as Torvalds embraces Sashiko for catching "embarrassing bugs" (E14) and highlights its 53.6% detection rate (E10), the kernel's own documentation still requires a human to review, license-check, and take full responsibility for any AI-assisted code (E2), and bars AI agents from the Signed-off-by tag that only a human can apply under the DCO (E7) — meaning the debate over Sashiko's workflow (E4, E6) unfolds within rules that already keep final accountability with people, not tools.

📊 Evidence

FAQ

What bug detection rate does Sashiko claim?

According to Tom's Hardware, Sashiko's project page states it can find 53.6% of bugs in proposed patches, with a false-positive rate pinned at within 20%.

Can an AI agent sign off on a Linux kernel patch?

No. According to ITHome, the kernel's AI-assisted development documentation prohibits AI agents from adding a Signed-off-by tag, since only humans can provide the legal certification required under the Developer Certificate of Origin (DCO). AI-assisted contributions instead require an Assisted-by tag naming the tool and model version.

Did Linus Torvalds always view AI tools favorably?

No. According to Tom's Hardware, Torvalds dismissed AI tools as overhyped back in 2024, but has since stated that AI is 'clearly a useful' tool, a view he says was not as clear even a year earlier.

E
EffectStory 編輯部Editorial Team

Related

BRIEF

AI Optical Transceiver Market Poised to Jump 57% in 2026 as Industry Pivots to LPO and Silicon Photonics

According to TrendForce, the global AI optical transceiver market is forecast to grow from $16.5 billion in 2025 to $26 billion in 2026, up more than 57%, as North American hyperscale data center traffic keeps expanding above 30% a year. Morgan Stanley separately projects co-packaged optics (CPO) revenue rising from $8 million in 2023 to $9.3 billion by 2030, with NVIDIA-linked designs reaching 75% of that demand by 2027, even as TrendForce flags tight EML and CW-LD chip supply.

EffectStory 編輯部 ·
BRIEF

Sunda Technology's Board Approves Jufeng Merger; TrendForce Analyst Sees Wider Starlink Order Scope

According to a CNA (Central News Agency) report also carried by UDN, Sunda Technology's (昇達科技) board approved a merger with Jufeng Technology (巨頻科技) on July 18, 2026. TrendForce analyst Wang Wei-ju (王偉儒) said the deal could widen order scope from Sunda's existing low-earth-orbit satellite clients, while chairman Chen Shu-min (陳淑敏) cited ambitions to enter the AI data center supply chain.

EffectStory 編輯部 ·