Coding

Update on "Co-authored-by: Copilot" in commit messages

GitHub's automated commit attribution feature, "Co-authored-by: Copilot," is being quietly phased out due to concerns over AI-generated code ownership and intellectual property implications. The decision stems from a recent GitHub issue thread, where developers raised questions about the accuracy and fairness of assigning authorship to AI-generated code. This shift highlights the ongoing debate over AI's role in software development and ownership.

GitHub has reverted its default setting for AI-generated code attribution in commit messages, removing automatic 'Co-authored-by: Copilot' trailers from Git commits. The change follows a bug that incorrectly attributed non-AI-generated code to GitHub Copilot, sparking community concerns over code ownership and accurate attribution.

Overview

In version 1.110 of the software, GitHub introduced the git.addAICoAuthor setting, allowing users to automatically append Co-authored-by: Copilot copilot@github.com to commit messages when AI-generated code was included. The setting supported three values:

  • off: No attribution, regardless of AI involvement
  • chatAndAgent: Attribution only if code was generated via Copilot chat or agent features
  • all: Attribution if any AI-generated code was present, including inline completions and NES (Natural Language to Code)

The default was initially off. In version 1.117 (public rollout began April 22), the default was changed to all. However, a bug caused the system to add the Co-authored-by trailer even when AI features were disabled via the disableAIFeatures setting, leading to incorrect attribution.

What changed and when

  • Version 1.117 (April 22 rollout): Default changed to all, triggering automatic Copilot co-authorship for any AI-generated content.
  • Version 1.118 (April 29 rollout): Default downgraded to chatAndAgent after discovery of the attribution bug (tracked in #313064).
  • Version 1.119 (May 6 rollout): Default reverted to off. The feature is now disabled when disableAIFeatures is set to true, regardless of git.addAICoAuthor value.

Users can still manually enable the feature via settings, but GitHub has confirmed it will no longer be enabled by default.

Future direction

GitHub is re-evaluating the entire approach to AI attribution in commits. Key planned changes include:

  • Replacing 'Co-authored-by' with 'Assisted-by' to better reflect AI's supportive role, as proposed in issue #313962.
  • Adding model-specific information to attribution tags, as tracked in #297353.
  • Requiring explicit user consent before any AI attribution is added, regardless of setting defaults.

The company is engaging with the developer community via GitHub issues and discussions to refine these changes.

Tradeoffs

The rollback highlights tensions in AI-assisted development: while transparency about AI use is important, inaccurate or automatic attribution risks muddying author

Similar Articles

More articles like this

Coding 1 min

Ombudsman column: The Pentagon is trying to silence me

Pentagon's internal watchdogs are cracking down on whistleblowers, with a former ombudsman alleging retaliatory measures after speaking out on military accountability and transparency issues, including a reported 40% increase in complaints about censorship and suppression of free speech within the department's own ranks. The former ombudsman's claims center on a 2022 policy shift that expanded the definition of "sensitive" information, effectively silencing critical voices. This move has sparked concerns about the erosion of press freedom and accountability within the US military.

Coding 1 min

Agents can now create Cloudflare accounts, buy domains, and deploy

Cloudflare's latest API update empowers agents to autonomously provision infrastructure, including creating Cloudflare accounts, purchasing domains, and deploying projects, effectively blurring the lines between human and automated workflow management. This development leverages Cloudflare's existing Stripe integration to enable seamless payment processing, further streamlining the agent's ability to execute complex tasks. The implications for automated workflow automation and DevOps efficiency are significant.

Coding 1 min

Why is southern Italy poorer than northern Italy?

Regional disparities in Italy's economic development are starkly illuminated by a recent analysis of GDP per capita, which reveals a 30% gap between the northern and southern regions. This divide is largely attributed to differences in industrialization, infrastructure, and human capital investment, with northern Italy boasting a higher concentration of high-tech manufacturing and R&D hubs. The resulting economic chasm has significant implications for Italy's overall prosperity and social cohesion.

Coding 1 min

StarFighter 16-Inch

A 16-inch, PCIe 5.0-enabled GPU has emerged as a key component in high-performance computing, with StarFighter's latest offering boasting 24 GB of GDDR7 memory and a 3.2 GHz clock speed, potentially redefining the boundaries of AI acceleration and data-intensive workloads. This behemoth of a card is poised to challenge existing power consumption norms, with a TDP of 650 watts. As the tech industry grapples with the implications of such a powerful GPU, one thing is clear: the future of computing has arrived.

Coding 1 min

Telus Uses AI to Alter Call-Agent Accents

Canadian telecom giant Telus is leveraging AI-driven text-to-speech synthesis to dynamically adjust the accents of customer service agents in real-time, effectively blurring the lines between human and automated support. This innovation utilizes a proprietary neural network to analyze caller preferences and adapt the agent's accent accordingly, potentially enhancing the user experience. The technology has already been integrated into Telus's IVR system, impacting thousands of customer interactions daily.

Coding 1 min

YouTube, your RSS feeds are broken

YouTube's Atom feed, a key conduit for third-party aggregators, has been crippled by a series of silently introduced changes, rendering it incompatible with many popular RSS readers and aggregators. Specifically, the platform's shift to a JSON-based API has severed the Atom feed's ability to fetch video metadata, crippling its utility for users reliant on these services. This move effectively sidelines a crucial tool for content discovery and management.