Using the open-weight model GLM 5.3, researchers created a malicious webpage that silently extracted local files from machines running vulnerable Mozilla Firefox versions. The automated attack bypassed obfuscation tactics meant to disguise patches as normal maintenance and even discovered an unpublished flaw to succeed. While attempts against Google Chrome were unsuccessful, the experiment cost only around $1,000 per browser version without triggering safety guardrails. Mozilla has since addressed the critical flaws with the release of version 159.0.
Key points
Publicly visible code updates make open-source software particularly vulnerable to automa…: Publicly visible code updates make open-source software particularly vulnerable to automated patch-gap exploitation by open-weight AI systems.
The experiment showed that AI models can bypass patch obfuscation efforts and locate supp…: The experiment showed that AI models can bypass patch obfuscation efforts and locate supplementary, unpublished flaws without triggering provider safeguards.
Software projects with public fix pipelines face emerging security threats as the cost an…: Software projects with public fix pipelines face emerging security threats as the cost and technical expertise required for automated exploitation decline.
Sources checked by our analysis pipeline: 1
Overview
GLM 5.3, a widely available open-weight model released by Z.ai in August, can identify and exploit vulnerabilities in recent versions of the Mozilla Firefox browser using security fixes that have been published but not yet shipped to users. The window between a fix being published and reaching users is known as the patch gap.
In our experiments, GLM 5.3 used this gap to build a webpage that, when visited by a Firefox user, read a file from the visitor’s computer and sent it to the attacker with no further user interaction. In particular, we found that any Firefox version released between September 17 and October 26 could be compromised in this way using only information available at the time of release.
The patch gap is a known weakness in software release practices, and it exists widely across software, e.g., browsers, phone operating systems, and router firmware. Exploiting it previously required significant time and expertise, or special access to frontier models with advanced cyber capabilities, so few attackers could target these vulnerabilities before they were fixed.
Rapid integration of automated tools across the state's legal sector has failed to deliver widespread cost savings, as practitioners spend significant unbilled time verifying outputs and fixing errors. Practitioners frequently find that clients using external chatbots independently generate flawed research and documents that require time-consuming correction. Meanwhile, judicial authorities are intensifying regulatory oversight following incidents involving fabricated case citations.
The chipmaker's potential financial backing aims to support the debt required to construct a massive compute facility for an emerging AI cloud provider. This arrangement would directly expand the physical infrastructure available to power intensive model development. Ultimately, the partnership positions Anthropic as the primary tenant to take advantage of the newly developed capacity.
Higher education coverage often fixates on rising tuition costs and career outcomes in a rapidly evolving job market. However, long-standing interpersonal capabilities taught on campus are becoming increasingly essential as artificial intelligence reshapes employment. Equipping graduates with these human interactions helps ensure they remain competitive alongside automated technologies.