GLM 5.3 exploits Firefox patch gap to steal local files
Epoch AI reports in a research publication that Z.ai's open weight GLM 5.3 model used published security fixes to exploit recent Firefox versions before patches reached users. In tests, the model created a webpage that could read a local file and send it to an attacker when a Firefox user visited it, without further interaction.
The researchers compromised every Firefox version released from September 17 to October 26 using information available when each version shipped. Once the testing setup was complete, each successful attack cost about $1,000. GLM 5.3 identified security related code changes despite efforts to present them as ordinary improvements.
The attacks also used an unpublished vulnerability alongside flaws identified through the patch gap. Epoch AI disclosed the issue to Mozilla, which fixed the unpublished vulnerability in Firefox 159.0 on October 27. The researchers said this change should make similar attacks more difficult.
We hope you enjoyed this article
Consider subscribing to one of our newsletters like Cybersecurity AI Weekly or Daily AI Brief.
Also, consider following us on social media:
More from Cybersecurity
Oct 30 Cursor Removes Malicious Claude Code Extension, Warns Up to 5,000 Users Oct 30 ECCC Launches €42 Million AI Cybersecurity Project Cluster Oct 30 Canada's Biggest Banks Block AI Agents From Customer Accounts Oct 30 Attacker impersonates 32 AI bots to hunt for credentials Oct 29 Hugging Face Malware Campaign Returns After TakedownCybersecurity AI Weekly
Weekly newsletter about AI in Cybersecurity.
Industry analysis
2025 Global Business Services Agenda: Gen AI Takes Center Stage
This industry analysis by The Hackett Group explores the transformative impact of generative artificial intelligence (Gen AI) on global business services (GBS) in 2025. The study highlights the shift from exploration to acceleration of Gen AI initiatives, with 89% of executives advancing these projects to improve customer satisfaction, innovate products, and reduce costs. The report also discusses the challenges and strategies for successful Gen AI adoption, emphasizing the need for a technology-enabled operating model and the importance of reskilling the workforce.
Read moreYou may also like
Anthropic Model Builds Exploits for Chrome, Safari, Windows and Linux
Cloudflare Bug Exposed Customer Data Across Workers AI Requests
XBOW Builds Chrome Exploit Chain With OpenAI Model in Six Days
OpenAI Coordinates Fixes for Critical PHP Vulnerabilities
Eclypsium Shows Hacked AI Hardware Can Alter Chatbot Answers
Daily AI Brief: the AI news that matters, in your inbox.