A critical flaw in TianoCore EDK II, the open source code that hardware makers and cloud providers use to build the firmware that runs when a server starts, was credited to two researchers from the AI labs Anthropic and Google. The CVE record for CVE-2026-97631, published on 22 October 2026, names Alex Matrosov (Anthropic) and Dan Holmes-Mitra (Google) as the finders and gives it a CVSS 4.0 base score of 9.5, rated critical.
TianoCore’s own advisory thanks the two “for finding this issue and providing the recommended fix.” Ars Technica reported on 27 October that the bug had leaked before a coordinated disclosure that was planned for January.
Short explainer · Security brief
A vertical news Short narrated from the published headlines and opening lines of this piece and five related FourWeekMBA security pieces.
Business Pill · Embedded dependency
An evergreen Business Engineer explainer on what sits inside the systems you depend on. It explains the general idea and does not discuss this flaw.
The key insight: As we read it, the sources tie this flaw to two AI labs only through who found it: the CVE record credits researchers at Anthropic and Google, and none of the sources says AI tools were used.
Update (28 Oct): Linux distributions list the flaw as unfixed. The Debian security tracker marks its edk2 package as “vulnerable” in every listed release (bookworm, trixie, forky and sid), with the fixed version for unstable shown as “(unfixed)”; for trixie it notes “no-dsa” and “(Minor issue)”. It links the three upstream EDK II commits as “Fixed by”.
Red Hat’s CVE database rates the flaw “Critical” and lists the edk2 package in Red Hat Enterprise Linux 8, 9 and 10, plus two OpenShift Container Platform 4 components, as “Affected”.
Ars Technica also reported that Project Mu, Microsoft’s hardened EDK2 fork, “used the code containing the vulnerable component as well.” As of this update, we found no firmware update from a server maker that names the CVE.
The advisory, titled “Pre-boot RCE in EDK II PXE driver,” says the flaw sits in the driver that lets a machine boot over a network. It says an unauthenticated attacker on the network could, while a machine is network-booting, “execute arbitrary code before the operating system loads, before Secure Boot or measured boot has verified or measured any OS component.”
It says “the exposure window is the duration of the PXE boot,” the network-boot step. The CVE record lists EDK II’s NetworkPkg as affected in versions up to and including edk2-stable202608.
This piece describes the flaw only at the level the advisory and the reporting state. It does not cover how an attack would work.
CVE-2026-97631 from ID reservation to public advisory. The CVE ID was reserved on 24 September; a public pull request with a fix, which Ars Technica says AMI published, opened on 28 September; fix pull request #13302 was opened and merged on 14-15 October; the advisory and CVE went public on 22 October with a CVSS 4.0 score of 9.5. Ars Technica reported on 27 October that disclosure had been planned for January. Sources: CVE record, TianoCore advisory, GitHub, Ars Technica.
What Was Posted on X
โผ๏ธ Anthropic and Google researchers found a flaw that lets one crafted packet hijack a server during network boot, before Secure Boot checks the OS.
Almost all servers on earth are affected.
CVE-2026-97631 sits in TianoCore EDK II, code in the UEFI firmware of most servers.
— International Cyber Digest (@IntCyberDigest) Oct 28 05:50
International Cyber Digest posted on X that Anthropic and Google researchers found the flaw, CVE-2026-97631 in TianoCore EDK II, and wrote that almost all servers on earth are affected; the piece sets out the scope above.
Who Is Exposed, per Ars Technica
Ars Technica described EDK II as “boot-time firmware used by most of the world’s server vendors and data centers.” It also said the bug “doesn’t affect home and small office users, or any other users who boot their devices solely from disk drives,” and applies only where firmware uses EDK II and network booting.
Ars reported that “some level of control over a local network is a prerequisite” for exploitation, and that “there’s no evidence yet that CVE-2026-97631 provides an initial entry point into a network.”
Intel told Ars that none of its “commercial products currently under security support” are known to be affected, and that “Intel does not include the affected EDK2 functionality in its reference code, though it may be integrated by an OEM board vendor or system manufacturer.” Arm told Ars: “CVE-2026-97631 is not directly applicable to Arm’s product portfolio.” Ars said questions sent to AMD, Phoenix Technologies, AMI, Microsoft and Amazon had gone unanswered when it published.
How It Went Public Early
From the public pull request with a fix (28 September, GitHub) to TianoCore’s advisory (22 October): 24 days by our count. The advisory and CVE record give a CVSS 4.0 score of 9.5, critical. Sources: TianoCore advisory GHSA-99cj-7hg9-mp4m; CVE-2026-97631; GitHub pull request #13200.
Ars reported that the two finders “had arranged for it to stay under wraps until a coordinated disclosure that was scheduled for January,” and that AMI published a pull request detailing the bug and a fix the month before. GitHub shows that pull request, #13200, opened on 28 September. TianoCore’s advisory went public on 22 October, 24 days later by our count.
A separate fix, pull request #13302, was opened on 14 October and merged on 15 October, GitHub shows. The advisory names #13302 as the patch, while its patched-versions field still reads “Not patched currently.”
What Fixing It Takes
Ars wrote that “patching UEFI code vulnerabilities in millions of end devices sold and used by thousands of hardware makers and cloud providers is cumbersome and time-consuming.”
Jesse Michael of the firmware security firm Eclypsium told Ars that “this vulnerability can be fixed with a basic code fix, which will require deploying a BIOS/UEFI update.” Ars reported that HD Moore of Dragos said the bug is likely a “forever” bug that will live on in end-of-support systems.
As we read it, the clock started early: Ars Technica says disclosure was planned for January, while GitHub shows a public pull request with a fix on 28 September.
As we read it, the exposure is narrower than the code’s reach: Ars says the bug applies only to organizations whose machines boot from networks and whose firmware uses EDK II, and that exploitation needs some control over a local network.
As we read it, the fix and the rollout are separate steps: GitHub shows the fix merged on 15 October, and Eclypsium’s Jesse Michael told Ars it will require deploying a BIOS/UEFI update.
TianoCore EDK II advisory, 22 October 2026
“Thanks to Alex Matrosov (Anthropic) and Dan Holmes-Mitra (Google) for finding this issue and providing the recommended fix”
Three Implications
FOR SERVER MAKERS AND FIRMWARE VENDORS Ars reports that questions to AMD, Phoenix Technologies, AMI, Microsoft and Amazon went unanswered when it published, and that patching UEFI code across many devices is “cumbersome and time-consuming.”
FOR DATA-CENTRE AND IT OPERATORS Ars reports the bug applies only to organizations with machines that boot from local networks, or conceivably from Internet-connected resources, and says operators should check with their firmware maker and consider turning off network boot temporarily if possible.
FOR AI LABS The CVE record credits researchers employed by Anthropic and Google as the finders of a critical flaw in widely used firmware code.
The framework puts it this way: “The value proposition shifts from features to integration depth.”
As we read it, boot firmware is integration depth at the lowest layer: Ars describes EDK II as code that hardware makers and cloud providers build on, so one flaw reaches many products and a fix has to travel through each of them. The framework was written about enterprisesoftware and does not discuss firmware or this flaw.
What Is Not Established
The CVE record, the advisory and the Ars Technica report we read do not say whether AI tools were used to find the flaw, and we do not claim they were. They name individual researchers and their employers.
They do not give a count of affected servers or a list of vendors that have shipped firmware updates. Ars said answers from several large vendors were not available when it published.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
The CVE record credits researchers at Anthropic and Google with a critical network-boot flaw in firmware code that Ars Technica says most server vendors use. Ars reports it went public before a disclosure planned for January, and Eclypsium’s Jesse Michael told Ars the fix will require deploying a BIOS/UEFI update.
A note on sourcing. The score, affected versions, credits and dates come from the CVE record, TianoCore’s advisory and GitHub. Scope, the early leak, vendor statements and expert quotes are Ars Technica’s reporting, attributed as such. This piece stays at the level of what is affected and how it is being fixed; it does not describe how an attack would work.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
