Skip to content
Misuse1 sourceFirst seen
Picture from the epoch.ai article
Image: epoch.ai

Open-weight model reads the patch notes nobody was supposed to read yet, then lets itself into Firefox

Ourtake

Security fixes become public before users get them, and it turns out that's a very long hallway for a robot with a $1,000 budget.

What happened

Researchers found that GLM 5.3, a widely available open-weight model from Z.ai, could use security fixes published before release to build working exploits against six recent Firefox versions. The exploits let a visited webpage read a local file without user interaction, at roughly $1,000 per version. Mozilla has since shipped Firefox 159.0, which patches the unpublished vulnerability the attacks also relied on.

Source