Hugging Face says an AI agent carried out an end-to-end cyberattack

Hugging Face reported a cyberattack executed end-to-end by an autonomous AI agent system, marking a shift from AI-assisted to AI-led hacking. The AI agent performed thousands of automated actions, exploiting vulnerabilities and stealing credentials over a weekend. Interestingly, AI also assisted Hugging Face in detecting and reconstructing the intrusion, highlighting the evolving landscape of AI in cybersecurity.

• Hugging Face experienced an internal database breach driven by an autonomous AI agent.
• This incident is one of the first documented cases of an AI agent conducting an entire cyberattack.
• The AI agent executed tens of thousands of automated actions, uploaded malicious data, exploited vulnerabilities, and stole credentials.
• Hugging Face used AI, specifically GLM-5.2, to analyze the malware and reconstruct the attack after initial frontier models were blocked by safety guardrails.
• The company is investigating potential access to customer or partner datasets and has not publicly attributed the attack.
• The incident highlights the growing trend of AI-powered cybercrime and the future of AI in both attack and defense.
  Continue reading https://www.axios.com/2026/07/20/hugging-face-ai-cyberattack-data-breach