Built on Stolen Data

The author argues that GitHub's Copilot, released as a VSCode plugin, was trained on restrictively licensed code and that this makes its use and distribution deeply immoral and illicit. The article cites a tweet by eevee saying GitHub admitted Copilot was trained on large amounts of GPL code, and a tweet by Armin Ronacher showing Copilot reproducing the fast inverse square root algorithm from Quake III (GPL licensed). A July 8, 2021 edit quotes Nora Tindall saying GitHub Support confirmed by email that it used all public GitHub code for Codex/Copilot regardless of license.