A model-written browser exploit is caught in the wild by June 30th 2027
Why I called it
Epoch AI showed an model turning each new Firefox release's public fixes into a working exploit for about $1,000, and state groups already run open models in their operations. If the cost of an exploit has fallen this far, an in-the-wild case attributed to a model should surface within eight months. If none does, the was not the constraint attackers faced.
The call, in full. A browser vendor, CISA or a named security firm reports in-the-wild exploitation of a Chrome, Firefox, Safari or Edge vulnerability using an exploit it says was built with an AI model, by June 30th 2027.
Scoring criterion. RESOLVES CORRECT if between 2026-11-01 and 2027-06-30 23:59 UTC Google, Mozilla, Apple, Microsoft, CISA, or a named security research firm (for example CrowdStrike, Mandiant, Cisco Talos, Wiz, Kaspersky, ESET or Microsoft Threat Intelligence) publishes that a vulnerability in Chrome, Firefox, Safari or Edge was exploited in the wild using an exploit the publisher states was developed or substantially developed with an AI model. RESOLVES WRONG otherwise.
The criterion is the machine-checkable version: a prediction that cannot be settled by a third party against a public source fails the build before it reaches this page.