voice-biometrics data-breach ai-contractors cybersecurity

4TB of voice samples just stolen from 40k AI contractors at Mercor

4TB of voice samples were just stolen from 40,000 AI contractors

app.oravys.com

April 27, 2026

6 min read

🔥🔥🔥🔥🔥

69/100

Summary

4TB of voice samples from 40,000 AI contractors were stolen and posted by the extortion group Lapsus$ on April 4, 2026. The leaked data includes voice biometrics linked to government-issued identity documents, raising concerns about potential misuse.

Key Takeaways

4TB of voice samples from 40,000 AI contractors were stolen by the extortion group Lapsus$ and posted on their leak site on April 4, 2026.
The breach combines voice biometrics with government-issued identity documents, enabling potential misuse for identity theft and fraud.
Attackers can use the stolen data for various malicious activities, including bypassing bank verification, committing vishing scams, and executing deepfake video calls.
Individuals who uploaded voice samples to Mercor should treat their voice data as compromised and take steps to protect their identity.

Read original article

View story →

Coverage & discussions

Hacker News · app.oravys.com

4TB of voice samples just stolen from 40k AI contractors at Mercor ↗

Community discussions

Hacker News discussion ↗

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Negative

Positives

The offer from ORAVYS to analyze suspect voice samples for free highlights a proactive approach to addressing potential misuse of AI-generated voice data.
The discussion around 'Datensparsamkeit' emphasizes the importance of data minimization, which is crucial for protecting individual privacy in AI applications.
Concerns about the origins of training data for text-to-speech models could lead to greater transparency and ethical standards in AI development.

Concerns

The situation with Mercor contractors reveals a troubling pattern of consent manipulation, raising serious ethical concerns about how AI companies handle personal data.
The reliance on biometrics as a form of identification is increasingly seen as a liability, especially when users are unaware of the permanence of their data.
The lack of clarity regarding the training data for current text-to-speech models suggests a significant gap in ethical AI practices, potentially undermining trust in the technology.