PyTorch-nightly Compromised via Dependency Confusion on PyPI
PyTorch-nightly
Attackers uploaded a malicious dependency package to PyPI mimicking PyTorch-nightly, causing pip to install compromised code. The supply chain incident exposed sensitive information from affected Linux machines.
- fix status
- Fix status unknown
- first report
- 3 Feb 2022
- last update
- 3 Feb 2022
Timeline · 3
- 3 Feb 2022Compromised PyTorch Dependency Chain
- 3 Feb 2022PyTorch statement on compromised dependency
- 3 Feb 2022Analysis by BleepingComputer
Sources · 3
News and write-ups
- 2022-02-03PyTorch statement on compromised dependencypytorch.org · pytorch.org/blog/compromised-nightly-dependency
- 2022-02-03Analysis by BleepingComputerBleepingComputer · bleepingcomputer.com/news/security/pytorch-discloses-malicious-dependency-chain-
Research and catalogues
- 2022-02-03Compromised PyTorch Dependency ChainMITRE ATLAS · atlas.mitre.org/studies/AML.CS0015
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Supply chain
- AI role
- Attack on AI
- categories
- Supply chain
- severity
- 3 · Confirmed exploitation
- fix status
- Fix status unknown
- targets
- Framework / tooling
- owasp llm top 10
- LLM03:2025 Supply Chain
- mitre atlas
- AML.T0010 AI Supply Chain Compromise
- dates
- happened 3 Feb 2022 · first reported 3 Feb 2022
- origin
- MITRE ATLASDerived from MITRE ATLAS (Apache 2.0, © MITRE).
- record
- confidence 0.95 · created 2026-10-18 06:55Z · updated 2026-10-18 06:55Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# PyTorch-nightly Compromised via Dependency Confusion on PyPI https://gcdradar.vercel.app/incidents/pytorch-nightly-compromised-via-dependency-confusion-on-pypi · AIR-2022-0031 · closed 2022-02-17 - Lane: Supply chain - Severity: 3 (Confirmed exploitation) - Fix status: Fix status unknown - First reported: 2022-02-03 - AI role: Attack on AI - Products: PyTorch-nightly ## Summary Attackers uploaded a malicious dependency package to PyPI mimicking PyTorch-nightly, causing pip to install compromised code. The supply chain incident exposed sensitive information from affected Linux machines. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2022-02-03: Analysis by BleepingComputer (https://bleepingcomputer.com/news/security/pytorch-discloses-malicious-dependency-chain-compromise-over-holidays) - 2022-02-03: PyTorch statement on compromised dependency (https://pytorch.org/blog/compromised-nightly-dependency) - 2022-02-03: Compromised PyTorch Dependency Chain (https://atlas.mitre.org/studies/AML.CS0015) ## Sources - 2022-02-03 MITRE ATLAS: Compromised PyTorch Dependency Chain https://atlas.mitre.org/studies/AML.CS0015 - 2022-02-03 pytorch.org: PyTorch statement on compromised dependency https://pytorch.org/blog/compromised-nightly-dependency - 2022-02-03 BleepingComputer: Analysis by BleepingComputer https://bleepingcomputer.com/news/security/pytorch-discloses-malicious-dependency-chain-compromise-over-holidays Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closedPyTorch-nightly PyTorch-Nightly PyPI Package Compromised with Malicious Dependency
Supply chain · Exploited · since 25 Dec 2022
AI - closed9.0Hugging Face libraries CVE-2023-6730: Deserialization vulnerability in Hugging Face transformers
Patch now · Disclosed · since 19 Dec 2023 · CVE-2023-6730
AI - closedHugging Face libraries PoisonGPT Open Source Model Supply Chain Demonstration
Research · since Jul 2023 (approx.)
AI - closedMobile deep learning apps Backdoor Attacks on Deep Learning Models in Mobile Apps
Research · since 18 Jan 2021
AI - closedHugging Face libraries AI Hallucinations Exploited for Supply Chain Attacks via Slopsquatting
Breaches & attacks · Exploited · since 1 Dec 2023
AI - closedMobile Deep Learning Apps Mobile Deep Learning Apps Vulnerable to Neural Payload Injection
Research · since 31 Mar 2023
AI