Linux Kernel Set to Retire Mandatory AI Disclosure Tag on Patches
Linux kernel developers are moving to scrap the mandatory Assisted-by: trailer that has, until now, been expected on patches created with the help of AI or large language model tools. Phoronix reported on 11 August 2025 that the proposal has drawn enough support on the kernel mailing lists for maintainers to treat the requirement's removal as a matter of when, not whether.
The change is still a proposal in progress — nothing is final until it lands in the kernel's documented contribution guidelines and maintainers adjust their trees accordingly. But the direction of travel is clear, and it has practical consequences for anyone shipping code upstream.
What the tag was for
The Assisted-by: trailer was introduced as a lightweight transparency measure. If an AI coding assistant, autocomplete plugin, or LLM chatbot materially shaped a patch, the submitter was expected to attach a trailer naming that tool, much as Co-developed-by: and Acked-by: trailers already record other forms of involvement in a change.
The intent was sound: reviewers wanted visibility into how a patch was produced, particularly as LLM-generated code began flooding mailing lists with confident-looking but under-tested contributions.
Why it is being dropped
The problem is that the signal evaporated. As AI assistance became embedded in editors, IDEs, and command-line tools, almost every patch touched an LLM somewhere in its lifecycle. A disclosure tag that applies to nearly everything carries no discriminating power — it becomes noise that maintainers must read past rather than information they can act on, and a requirement that cannot meaningfully be enforced.
Kernel developers are effectively betting on review rather than disclosure as the quality gate. Under that logic, a patch either survives maintainer scrutiny or it does not — regardless of whether a human or a model drafted it. Checkbox-style transparency is being traded for the older, harder discipline of code review.
What is not changing
It is important not to over-read the shift. The Developer Certificate of Origin and the Signed-off-by: chain remain fully in force. Every contributor still attests that they have the right to submit the work and that it is their responsibility. Removing the AI tag does not create an anonymous contribution pipeline; it removes one annotation from an already accountable process.
Nor does the trailer disappear. Developers who want to record AI assistance can continue to do so voluntarily. What changes is compulsion, not availability.
What this means for upstream contributors in Hong Kong and Macau
For engineers in Hong Kong and Macau who maintain or contribute upstream patches with AI-assisted tooling, the immediate workflow impact is small: no trailer needs to be drafted or justified before sending a patch to a maintainer. The longer-term question is internal. Teams that wrote local engineering policies around AI disclosure — mirroring the kernel's tag — now have a live precedent for relaxing them, but those policies were rarely solely about upstream compliance. They usually also served audit, IP, and code-quality purposes that the kernel's decision does not address.
The bigger open question is what other large projects adopt in place of mandatory disclosure. The kernel is removing a requirement; other ecosystems may move the opposite way, which would leave contributors navigating divergent expectations across projects.
The bottom line
The kernel community is treating pervasive AI use as a solved disclosure problem — the answer being that disclosure no longer carries information. Accountability now rests entirely on the review process and the existing sign-off chain. For contributors working across organisations, that shift in upstream expectations is a reason to revisit whether internal AI-disclosure policies still match reality before the next patch is sent.
Linux 核心擬取消補丁強制 AI 披露標籤
Linux 核心開發人員正計劃取消強制要求的 Assisted-by: trailer，該標籤至今一直是借助人工智能或大型語言模型（LLM）工具製作的補丁所預期附上的。Phoronix 於 2025 年 8 月 11 日報道，此項提案已在核心郵件列表中獲得足夠支持，維護者普遍認為取消該要求只是時間問題，而非是否取消的問題。
此項變更目前仍屬提案階段——在正式納入核心成文的貢獻指南、並由維護者相應調整其分支之前，一切尚未定案。但發展方向已經明確，對所有向上游提交代碼的人士而言，均會帶來實際影響。
該標籤的用途
Assisted-by: trailer 最初以輕量級透明度措施的形式推出。如果某個 AI 編碼助手、自動補全插件或 LLM 聊天機械人對補丁有實質影響，提交者便須附上標籤，註明該工具名稱，正如 Co-developed-by: 及 Acked-by: trailer 已用於記錄其他形式的變更參與。
其用意無可厚非：審核者希望了解補丁的產生方式，尤其在 LLM 生成的代碼開始大量湧入郵件列表、以表面自信但未經充分測試的貢獻充斥之際。
為何取消
問題在於該信號已經失效。隨著 AI 輔助功能逐漸嵌入編輯器、IDE 及 command line 工具，幾乎每一個補丁在其生命週期中的某個環節都曾涉及 LLM。一個適用於幾乎所有情況的披露標籤並無任何區分作用——它只會淪為噪音，維護者須費神略過，而非可供他們採取行動的資訊，同時亦成為一項無法有效執行的要求。
核心開發人員實質上是選擇以代碼審核而非披露作為品質關卡。在此邏輯下，一個補丁若非經受得住維護者的審核，便是經受不住——不論是由人手抑或由模型起草。以打勾方式達致的透明度，正被換取回較古老、亦更艱難的代碼審核紀律。
不變的部分
切勿對此轉變過度解讀。《開發者出處證明》（Developer Certificate of Origin）及 Signed-off-by: 鏈仍然完全有效。每一名貢獻者仍須聲明其有權提交該項工作，並為此負責。取消 AI 標籤並不會造成匿名貢獻渠道；它只是從一個原本已需承責的流程中移除一項註釋而已。
該 trailer 亦不會消失。希望記錄 AI 輔助的開發者仍可繼續自願加上。改變的只是強制性，而非可用性。
對港澳兩地上游貢獻者的影響
對於香港及澳門利用 AI 輔助工具維護或向上游提交補丁的工程師而言，即時的工作流程影響甚微：在把補丁發送給維護者之前，無須再草擬或為任何 trailer 作說明。較長遠的問題則在於內部。部分團隊曾參考核心的標籤，制定關於 AI 披露的內部工程政策，如今有了放寬政策的現實先例，但這些政策往往並非純粹為了符合上游要求，通常亦兼有審計、知識產權及代碼品質等用途，而核心本次的決定並未觸及這些方面。
另一個更大的未知數是，其他大型項目會以什麼取代強制披露。核心是在移除一項要求；其他生態系統則可能反其道而行，令貢獻者須在不同項目之間應對各異的要求。
總結
核心社群現已將 AI 的普遍使用視為一個已解決的披露問題——答案就是披露已不再具備資訊價值。問責現在完全依賴審核流程及現有的簽署鏈。對於跨機構工作的貢獻者而言，上游期望的這項轉變，正是在下次提交補丁之前重新檢視貴機構內部 AI 披露政策是否仍然切合現實的理由。