Google researcher Matteo Rizzo reports a boot-time authentication bypass in AMD’s Zen 1 and Zen+ processors. CVE-2026-74127 lets attacker-controlled firmware run on the Platform Security Processor, the separate security core that starts before the main processor. AMD confirms the flaw and plans no mitigation because the vulnerable boot ROM is immutable hardware.
The attacker first needs access to modify the platform’s SPI flash, the chip storing its firmware. AMD explicitly warns that physical access is not always necessary. Whether software can install attacker-controlled firmware depends on the particular machine’s protections. This is not evidence that an ordinary network request can directly compromise every early Ryzen system.
Evidence status: researcher-demonstrated exploitation, acknowledged by AMD. The reviewed sources identify no confirmed campaign using this vulnerability against victims.
Firmware validation fails below the operating system
The Platform Security Processor normally verifies AMD-signed firmware before executing it. Rizzo describes a failure in the recovery path that lets previously loaded, unauthorized code survive a later successful signature check. The result is execution before normal platform startup, with access beyond the operating system’s visibility.
The research describes potential persistent malware with access to system memory and control of the main processor. Those are capabilities of the demonstrated compromise, not observations of an installed criminal implant or measured data theft.
An operating-system reinstall does not repair immutable boot ROM. It should not be treated as sufficient recovery after suspected firmware compromise.
The demonstrated machines and the affected family differ
Rizzo reports successful tests on Ryzen 5 1600X, a Zen 1 client processor, and Ryzen 7 2700X, a Zen+ client processor. The advisory says Zen 2 and later are unaffected.
The researcher had not tested a Zen 1 EPYC system and labels its susceptibility a likely extension of the earlier Masterkey research. AMD’s bulletin describes Zen and Zen+ processors as affected without providing a board-by-board exposure matrix. Keep the verified client demonstrations separate from the untested server case.
The GitHub advisory and AMD’s initial bulletin revision are dated November 17. The AMD-assigned CVE record was published November 21. That later catalog timestamp does not move the original disclosure date or establish when exploitation began.
Review firmware-write controls and hardware plans
For teams retaining these processor generations:
- Inventory the exact processor generation and motherboard. A Ryzen or EPYC brand name alone does not establish the affected architecture or its firmware-write protections.
- Ask the platform vendor how firmware writes are authorized and protected. AMD says this must be assessed per platform. Protect management access and firmware-update tooling as sensitive administration paths.
- Keep supported system and firmware updates current for other vulnerabilities. Neither source identifies a BIOS release that repairs this boot-ROM flaw. The advisory’s “Zen2” patched-version field describes a different hardware generation, not an installable patch.
- Include hardware replacement in high-assurance plans. Where policy requires resistance to attacker-controlled firmware, AMD offers no repair for the affected hardware. Prioritize according to the machine’s actual write exposure and role.
- Escalate suspected firmware compromise beyond routine desktop cleanup. Preserve available evidence and coordinate recovery with the hardware vendor and incident-response team before destructive reimaging.
Caveat: AMD confirms the defect, but the public evidence does not establish an attack campaign or a universal route past each motherboard’s firmware-write controls.