The report
Epoch researchers gave an open-weight model exploit tooling and information about a Firefox renderer weakness that was visible through pending public fixes. Against six tested Firefox versions, the resulting workflow constructed file-exfiltration chains by combining that clue with a distinct filesystem-protection bypass. Mozilla later patched the second weakness in Firefox 159.0.
The experiment matters because a renderer flaw and a working file-read chain are different outcomes. A standalone browser crash need not reach private files, but a second protection failure can change the effect. The report also supplies a missing cost denominator: a roughly one-thousand-dollar marginal run for an additional tested version does not include a six-figure initial research and setup effort. Comparable Chrome attempts did not yield success in the disclosed runs.
The Firefox chain dossier and Epoch study support lab capability under defined versions and tools, not a live mass-exploitation campaign. Response should prioritize updated browser builds and reproduce mitigation against the actual two-part chain. Treating the finding as either universal autonomous exploitation or merely an incorrect model-generated suggestion would discard what was specifically demonstrated.