Minitap Accuses Google's Artemis of Copying Its Code
A.I. / news
Minitap Accuses Google's Artemis of Copying Its Code
Google added a one-line credit to its Artemis repository after Minitap's co-founder showed matching commands, code and a shared bug from his startup's mobile-use project.

Google added a one-line attribution notice to its Artemis Android-automation repository after Minitap co-founder Luc Mahoux published commit-level evidence Sept. 11 that the project had copied his startup's open-source code and dropped the original authors' names.
Mahoux filed a GitHub issue the same day, writing that "Artemis includes code from our project, but we couldn't find an acknowledgement of Minitap, mobile-use, or its contributors." Minitap's own blog post, "I expected better from Google," laid out the specifics: matching Android device-connection code, word-for-word Hopper agent instructions, an identical WhatsApp messaging test scenario, and a shared bug in a media helper function.
What the commit history shows
Minitap's post said Artemis' pyproject.toml file originally carried mobile-use's version number, 3.6.3, and three names under a Google LLC copyright header: Pierre-Louis Favreau, Jean-Pierre Lo and Nicolas Dehandschoewercker, Minitap's chief executive. A force push in August replaced that file, removing all three names before Minitap began its own investigation in September. Both mobile-use and Artemis are licensed under Apache 2.0, which requires reusers to keep copyright and attribution notices intact.
Google's Artemis repository now carries a one-line notice in its license section: "This project includes source code developed by Minitap, Inc." Google has not otherwise responded to Mahoux's issue, which remained open with no reply as of Sept. 11.
One example from the post: both projects' messaging test scenarios script an agent to send identical Happy New Year greetings to contacts named Alice, Bob and Charlie. Minitap's post also traces a bug shared between the two codebases in a media-handling helper function, the kind of quirk Mahoux said is unlikely to appear twice by coincidence.
Two projects, one benchmark
Both tools compete on AndroidWorld, a Google benchmark that scores how reliably an agent completes tasks on a real Android device from natural-language instructions alone. Minitap says mobile-use reached 100% in its own evaluation in January. The benchmark's public leaderboard credited it with only 91.4% as of Sept. 11, a score last confirmed in December, while the same sheet listed Artemis at 99.1%, according to Minitap's post.
| Project | Maker | GitHub stars | AndroidWorld score, Sept. 11 |
|---|---|---|---|
| mobile-use | Minitap | 3,003 | 91.4% |
| Artemis | 5,839 | 99.1% |
Minitap raised $4.1 million in a seed round in December, co-led by Moxxie Ventures and Mercuri, with Hugging Face co-founder Thomas Wolf among the individual backers. Dehandschoewercker and Mahoux built mobile-use first; Google published Artemis on GitHub Aug. 13, saying it integrates with its own Antigravity, Codex and Claude Code coding assistants.
Google's handling of open-source terms has drawn separate scrutiny this month: its TimesFM-3 forecasting model shipped under a non-commercial license, a stricter change from the Apache terms on its two predecessors. Artemis itself joins a run of fast-growing open-source AI projects, including OpenClaw, whose governance took time to catch up with a sudden jump in stars.
Google has not said whether removing Minitap's names from Artemis' commit history was deliberate. Its repository still credits no individual engineer from either company, only the corporate line naming Minitap, Inc.
Sources
More in A.I.
- 01GPT-Rosalind-6 Reuses GPT-6 Astra's Weights, Cuts Biology Refusals to 9.4% and Stays Behind a GateOpenAI's Oct. 30 system card addendum says no training was added, so the difference from Astra is deployment controls and who is allowed in.
- 02Architect Says Claude Opus 5.5 Drew a House Set Good Enough for a Permit After Two Redline RoundsStephen Gallagher's informal test of 13 models gave Opus 5.5 the most human feedback, and the drawings still missed a required second exit.
- 0350 GPT Agents in Minecraft Built a Ration Ledger, and 34 Died Doing ItY Combinator's Labs group and Olam Labs ran GPT-6.1 Sol and Claude Opus 5.5 swarms through Minecraft and Mindustry. GPT won both head-to-head games, but each result comes from one harness and one setup.
- 04China's Announcement No. 48 Puts 23 AI Technologies on a Draft Export ListExporters of listed technology must already register with MOFCOM. Comments on the draft catalogue close on November 15, and the sources disagree on whether the thresholds are published.