- The Linux kernel community reached a consensus to deprecate the mandatory «Assisted-by» tag for AI-generated code.
- Linus Torvalds firmly rejected AI disclosure rules, arguing that author competence and code comprehension matter far more than the tools used.
- Contributors will no longer be forced to report whether advanced coding agents or LLMs assisted in preparing kernel patches.
The Linux kernel community has officially changed its stance on automated programming assistance. During the recent Maintainers Summit, core developers reached a consensus to eliminate the mandatory Assisted-by tag, which was previously required for any contribution generated with the help of artificial intelligence tools.
Sasha Levin, a Linux kernel developer and Nvidia engineer, confirmed the policy reversal on the Linux kernel mailing list, instructing maintainers to remove the requirement from official documentation.
Timeline of the AI Disclosure Policy
The concept of tagging machine-assisted contributions initially emerged in July 2025 and was formalized into kernel documentation in January 2026. However, rapid advancements in large language models made the regulation increasingly redundant.
| Date | Policy Status | Context |
|---|---|---|
| July 2025 | Initial Proposal | Early discussions on managing automated contributions |
| January 2026 | Mandatory Policy | Formal integration of the «Assisted-by» tag requirement |
| October 2026 | Deprecation | Maintainers Summit consensus to drop disclosure rules |
Why Kernel Maintainers Scrapped the Rule
Red Hat developer Jeff Layton outlined several fundamental reasons behind discarding the tag. First, compliance was inherently inconsistent. Second, appending specific model names turned commit logs into free promotional space for proprietary AI vendors. Third, the kernel project had no clear use case for collecting such metadata.
Additionally, the technical competence of code models has shifted. Speaking at the Linux Foundation’s Open Source Summit in South Korea, veteran maintainer Greg Kroah-Hartman noted that LLMs are now reliable enough to catch bugs because software code relies on structured patterns where modern pattern-matching models excel.
Torvalds: Understanding the Code Is What Matters
While some contributors, such as Ideas on Board founder Laurent Pinchart, argued that disclosure sets project expectations, Linux creator Linus Torvalds strongly rejected the mandate. Torvalds compared AI disclosure mandates to banning compilers or editors, pointing out that such rules fail to address actual code quality:
“The kind of expectations I have is that people do their best and I expect people to understand what they are doing. But that understanding has very little to do with what tools they use. And quite honestly, in my experience, the people I respect and call intelligent are not the ones that shy away from using new and interesting tools.”
Kernel maintainers will focus strictly on patch functionality, maintainability, and architectural fit rather than policing contributor workflows. Programmers can leverage AI tools freely, provided they fully understand and take complete responsibility for the code they submit.
While I respect Torvalds’ focus on code quality over tooling, dropping disclosures entirely feels short-sighted. LLMs introduce subtle, plausible-looking edge-case regressions that can easily bypass standard review. Without provenance metadata, auditing legacy patches when a foundational model issue is discovered down the line will be an absolute nightmare.