What it is
A Security describes its platform as an autonomous offensive-security and remediation system. Its public Zoomsday research used commercially available frontier models, reverse-engineering tools, and fewer than twenty prompts to map a proprietary protocol, identify memory-safety flaws, and build a working cross-platform exploit.
What is verified
The ledger counts CVE-2026-53413 and CVE-2026-53414. A Security’s disclosure documents the AI workflow, and Zoom’s bulletins corroborate the affected products and fixes. The operator supplies the AI attribution, so the entry is self-reported even though the vulnerability and remediation trail is public.
CVE-2026-53415 is deliberately excluded from A Security’s count. Its own write-up says Zoom had already found and fixed that independent flaw before A Security reported it.