🔍 Read the full analysis: Could Downloadable AI Change How We Handle Firefox Security Fixes? on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Epoch AI reports that downloadable model GLM 5.3 built attacks against six Firefox releases by analyzing publicly visible security fixes, with its fastest attempt taking 12 hours. Mozilla fixed an additional, previously unpublished flaw used in the tested attacks in Firefox 159, but Epoch has not yet tested whether the attacks work against that release.
Open-weight AI model GLM 5.3 built working attacks against six successive Firefox releases by analyzing public information about security fixes, according to a lab study published October 29 by research group Epoch AI. The attacks could make a browser read a local file when a user opened a webpage, the researchers said; Mozilla released Firefox 159 on October 27 with a fix for a separate, previously unpublished flaw used in the tests.
Epoch tested Firefox versions 156.0 through 158.0.1, released between September 17 and October 26. For each version, researchers gave GLM 5.3 access to information that was public at the time, including code changes that revealed security fixes not yet included in the stable release. Epoch calls the period before users receive and install a fix the patch gap.
The task was to build a webpage that could read a file from a computer running the specified Firefox version when the page was opened. Epoch says the model selected relevant fixes, developed an attack chain and dealt with failed attempts without further guidance. Each run had a limit of 48 hours; the fastest successful run took 12 hours. After the researchers built the test setup, they estimated API fees of about $1,000 per version. The overall setup and testing cost Epoch more than $100,000, according to its account.
The reported attacks also relied on a second vulnerability that had not been publicly disclosed. Epoch said GLM 5.3 rediscovered that flaw during each successful attempt and reported it to Mozilla. Mozilla fixed it in Firefox 159.0, released October 27. Epoch has not reported testing whether the same attacks work against version 159.
Infographic 01 · The patch gap
A published fix is a map. GLM 5.3 read it.
How an open-weight model turned Firefox security fixes into a working attack before the update reached users, in Epoch AI’s offline lab test.
Where the attack happens
The life of a security fix, simplified
- 1 · Fix writtenDevelopers find and fix a security bug
- 2 · Fix visibleThe change shows up in public code and test builds, disguised as an ordinary change. The patch gap opens
- 3 · Update shipsThe fixed version is released to everyone
- 4 · You install and restartOnly now is your browser protected. The gap closes for you
GLM 5.3 worked in step 2: it picked out the security fixes and turned them into a working attack before step 3.
Compromised in Epoch’s test
Released 17 Sep to 26 Oct 2026 · only information public at release
Result: a web page that read a local file and sent it to the attacker’s server when visited. No further user action.
Firefox 159.0 · 27 Oct
Fixes the unpublished flaw every recent success relied on
Epoch expects the fix to make similar attacks “at least substantially more difficult”.
Sources: Epoch AI, 29 Oct 2026; Firefox 159.0 release notes. Lab test without internet access; no real users were attacked. No exploit details are shown here.
Powered by Thorsten Meyer AI
Patch Details Can Speed Attack Development
The study points to a security risk in the time between a fix becoming visible and users installing the update. Publicly available code changes help developers and security researchers understand what has been repaired, but can also give attackers clues about how to target users who have not yet updated. Epoch’s results suggest that an open-weight model may help automate some of the analysis and development involved.
That does not establish that attackers are currently using GLM 5.3 against Firefox users, or that every published patch can be turned into an exploit. The study was a controlled exercise, and the tested attacks needed the additional unpublished flaw. Still, the speed reported by Epoch matters to software vendors and organizations: patch deployment delays may leave less time to respond if similar methods are used outside a lab.
For Firefox users, the direct practical step is to update to version 159 or later and restart the browser. For organizations, the report reinforces the value of applying browser security updates promptly. The findings concern a particular model, target and test setup, rather than all AI systems or all software vulnerabilities.
Products Worth Considering
Samsung Galaxy Tab S10 FE 128GB Wi-Fi Android Tablet, Blue
DJI Osmo Pocket 3 Creator Combo Vlogging Camera with 1 CMOS & 4K/120fps
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
How Firefox’s Patch Gap Was Tested
Firefox is an open-source browser, so changes to its code can be visible before a stable update reaches every user. Security fixes may also appear in beta or nightly builds. That visibility can reveal what kind of weakness developers addressed, creating a period in which the repair is known but some users remain on vulnerable software. Security researchers refer to this exposure as a patch gap.
Epoch says it supplied GLM 5.3 with the target Firefox version and basic tools, but no internet access. The test machine used Ubuntu 26.04, default settings and an official Firefox release. The model was instructed to examine fixes that had not yet reached the stable version. Researchers say they did not intervene during the runs. They chose file reading as the target because it could expose information such as browser login cookies, saved passwords or private documents.
Epoch also tried the setup against Chrome, but reported that it did not succeed after about $100,000 in attempts. It said older Firefox versions may also be vulnerable, but those releases were not systematically tested. The study’s results therefore apply to the versions and conditions Epoch examined, not to every Firefox installation.
“Widely available open-weight models can now do much, or even all, of this work on their own.”
— Epoch AI
Firefox 159 Has Not Been Retested
The study does not show whether the reported attack chain works against Firefox 159 or later. Mozilla fixed the previously unpublished flaw used in the successful tests, but Epoch said it had not yet evaluated the latest release and only expected the fix to make similar attacks substantially more difficult.
It is also unclear from the supplied study summary whether the method generalizes to other browsers, operating systems, or vulnerabilities. Epoch’s Chrome attempt failed after substantial spending, but that result does not establish that Chrome or other software cannot be attacked using AI. The researchers tested a specific file-reading goal in a controlled environment, not a live campaign against users.
The supplied material does not establish that real-world attackers have used GLM 5.3 in this way. Nor does it show how often a model could turn a public fix into a reliable attack under conditions outside the experiment. Epoch’s cost figures distinguish the more than $100,000 spent building and testing the setup from the roughly $1,000 in API fees it estimated for each subsequent Firefox version.
Update Firefox and Watch for Retesting
Firefox users should install version 159 or later and restart the browser, following Mozilla’s release guidance. Organizations that manage browsers across many devices can check that updates have reached those systems rather than relying only on an update being available.
The next technical question is whether the Mozilla fix prevents or materially disrupts the attack chain. Epoch said it had not tested Firefox 159 at the time of its report, so a retest or additional technical findings could clarify the patch’s effect. Mozilla’s release advisory identifies security fixes in the version, while the supplied material does not give a further public timeline for Epoch’s testing.
Researchers and browser vendors will also be watching whether the reported approach can be reproduced on other targets and whether faster patching reduces the exposure window. For now, the confirmed findings remain limited to Epoch’s controlled Firefox experiment and Mozilla’s release of a fix for the separate flaw disclosed by the researchers.
Key Questions
What did GLM 5.3 do in Epoch AI’s test?
Epoch says the model built webpages that could make specified Firefox releases read a local file when a user opened the page. The researchers conducted the work in a controlled lab environment, without internet access to target real users.
Which Firefox versions were tested?
The study reports successful attacks against six releases from 156.0 to 158.0.1, issued between September 17 and October 26, 2026. Epoch said older versions were not systematically tested.
Does the report show Firefox 159 is vulnerable?
No. Mozilla released Firefox 159 with a fix for a previously unpublished flaw used in the tested attacks. Epoch said it had not yet tested version 159, so the report does not confirm whether the attack works on that release.
What should Firefox users do?
Update to Firefox 159 or later and restart the browser. The study describes a lab result; it does not report a live attack on Firefox users.
Does the study prove downloadable AI can attack any browser?
No. The findings concern one model and a specific Firefox test. Epoch said its attempt against Chrome failed after about $100,000 in attempts, but that does not prove other browsers are immune or that the method applies broadly.
Source: ThorstenMeyerAI.com
