Feeds.Feedburner Researchers Compromise Hugging Face Account for Kokoro AI Model
Article Content
- •Researchers hijacked the Kokoro AI model account on Hugging Face.
- •The attack involved modifying model weights to produce altered outputs without detection.
- •Hugging Face does not verify publisher ownership, increasing supply-chain risks.
Security researchers gained control of the Kokoro text-to-speech model on Hugging Face, exposing vulnerabilities in AI model management. By registering an expired domain linked to the model's publisher, they reset the account password and modified the model's neural weights to produce altered audio outputs without changing the application code. This attack method allows for undetectable malicious changes, as traditional malware scanners cannot identify the compromise. Kokoro is widely used in projects by major companies like Google, NVIDIA, AMD, Intel, and Amazon Web Services. The researchers emphasized that Hugging Face lacks publisher ownership verification, enabling potential supply-chain attacks. The incident highlights significant trust issues in the open-model ecosystem, as even legitimate accounts could be exploited to distribute harmful content.
Ask AI about this cluster
Answers cite the sources they use
Timeline
More articles in this cluster (6)
Following this threat?
Track Amazon Web Services in your own feed — you're alerted when they show up in new reporting, leak sites or exploitation.
Free account · no card needed
Common questions
How did the researchers gain access?
What are the implications of this attack?
Is there a fix for this vulnerability?
Continue Reading
Lazarus Group Targets 30 South Korean Servers Using PHP Vulnerabilities North Korea-linked Lazarus Group has targeted 30 servers in South Korea as part of a larger cyber operation involving 581 servers across 10 countries. The attacks exploit two critical remote code execution vulnerabilities in PHP (CVE-2026-109618 and CVE-2026-109619), allowing unauthorized server command execution. The…