Artificial Intelligence

Researchers Hijack Hugging Face Account for Kokoro AI Model to Highlight Supply Chain Risks

· GlobeWire Daily Editorial Desk · Source: www.bleepingcomputer.com

2 min read

Key takeawaysSecurity researchers successfully took control of the Hugging Face account for the popular Kokoro text-to-speech model by registering an expired domain linked to the original developer and resetting the password. The demonstration aimed to expose a lack of ownership verification on the platform, which could allow attackers to distribute altered model…
Researchers hijack Hugging Face account for popular Kokoro AI model

Image: BleepingComputer · Photo source · Resized and letterboxed to 16:9; no AI generation

Security researchers successfully took control of the Hugging Face account for the popular Kokoro text-to-speech model by registering an expired domain linked to the original developer and resetting the password. The demonstration aimed to expose a lack of ownership verification on the platform, which could allow attackers to distribute altered model weights without triggering traditional malware scanners.

Security researcher Alon Hertz and his team demonstrated that modifying neural weights can alter the spoken output of the model without changing any underlying application code or configuration. The researchers stated that they tested the modification locally using a gaming GPU in 87 seconds but did not upload the poisoned weights to Hugging Face to avoid affecting production systems.

Key Findings and Methodology

  • Domain Registration: Researchers registered the domain listed in the project's public donation file to execute a password reset.
  • Stealth Modification: By altering only the model weights, the output changes while file sizes, delivery routes, and architecture remain consistent with legitimate releases.
  • Platform Vetting Issues: The team highlighted that Hugging Face currently does not verify whether accounts remain with their original owners or systematically vet uploaded models.

Practical Implications and Verification

  • Supply Chain Vulnerability: Systems that automatically fetch new revisions from open-source repositories could be exposed to modified outputs that lack traditional malware signatures.
  • Verification Steps: Organizations relying on open-weight models should independently verify publisher identities, pin specific model revisions rather than pulling latest updates blindly, and monitor integrity where feasible.
  • Response Status: Researchers reported the issue to Hugging Face but received no response at the time of reporting.
Share this article
Read the original source: Researchers hijack Hugging Face account for popular Kokoro AI model

AI-assisted reporting; source claims are attributed and have not been independently verified.

About the author

GlobeWire Daily Editorial Desk — AI-assisted news based on attributed published sources.

More to read