Anthropic’s Mythos unveil sent a shockwave through the cybersecurity community - but was this truly a new frontier in LLM capabilities to find 0days? While subsequent findings seemed to show these same bugs were discoverable by open weight models, these findings did not reflect real world conditions.
In this report, Xint’s award-winning security researchers demonstrate how the showpiece bugs found by Mythos were already discoverable using publicly available models wrapped in proper scaffolding that reflect real world offensive/defensive conditions.
