If you closed out your eMudhra cleanup in October and moved on to the next fire, your work may not be done after all. On November 7, eMudhra revoked eight more intermediate certificate authorities—the same eight it had assured Mozilla’s bug tracker had gone through a different code path and weren’t touched by the key compromise.

October’s eight had private keys weak enough that anyone could compute them from the public certificates. The fallout went far past the browser level when the first round of revocations hit, and eMudhra rapidly shut down its publicly trusted TLS business.

eMudhra has now posted an incident reportOpens a new window that expands what your cleanup has to cover. It also tells a fuller account of what happened. Among other things, the first warning sign now turns out to have surfaced back in June 2025. Given the upcoming remediation deadlines, it would be prudent to revisit the work you just completed if your audit relied on the vendor’s original attestations.

READ MORE: The von Neumann bottleneck: Why it still slows computers down

How one script took down 16 eMudhra CAs

If you want the gory technical details behind how this went down, here you go. Both waves of revocations trace to one key ceremony that was held on April 30, 2025. eMudhra needed new intermediate CAs for its U.S. and European facilities, but the hardware security modules (HSMs) ordered for the U.S. site were stuck in supply chain limbo. Root program policy changes meant the keys had to exist before May was out, however. So, the company generated 16 CA key pairs on the Utimaco HSMs it already had, planning to move them later to Crypto4A units at the new sites.

Vendors don’t make it easy to shift private keys between competing HSMs, and that’s by design. eMudhra’s operations team wrote a custom Python script to handle the move anyway. The ceremony included a rehearsal of the transfer, with the HSM’s FIPS-approved mode switched off as a sanctioned exception—the cryptographic equivalent of propping a vault door open for the movers.

The script wrapped each key for transport, unwrapped it, then pulled the result out of the HSM and reassembled a key object on the host machine. For the eight ECDSA keys, the report says, “a single line in that read-back truncated the private value to its leading byte.” A 48-byte secret shrank to a single byte.

Plenty of IT teams have a quick script that ended up with more responsibility than anyone meant to give it. This one was handling the private keys of a publicly trusted CA. The mangled copy carried a label nearly identical to the original’s, and no step in the ceremony checked key strength. The weak objects were certified, shipped off for cross-signing, and left to blow up in October.

Granted, the eight RSA keys kept their full length, and every signature still verifies. But their private components spent that rehearsal inside the memory of an ordinary Python process on the machine running the ceremony. eMudhra concedes it can’t prove that material stayed within the rules its own certification practice statement sets. Since no log can establish what happened inside that process, the company revoked all eight RSA intermediates on November 7. It used the reason code cessationOfOperation rather than keyCompromise. In certificate terms, that’s the label for an orderly wind-down, not for keys that can no longer be trusted. The choice is still drawing questions in the Mozilla bug thread.

What eMudhra’s all-clear was hiding

October’s revoked intermediates had issued no certificates of their own. The pain back then came from the trust-store fallout, not from subscriber certs. This batch is different. The revoked RSA issuers had produced 195 certificatesOpens a new window , 130 of them domain-validated TLS and 63 organization-validated, plus two for code signing.

READ MORE: The FAA ground stop started with a problem IT pros recognize all too well

The holders include several universities, a county, and 29 of eMudhra’s own domains. Somewhere, an IT office just finished replacing certificates it had every reason to think were fine. (The newest certificate was issued November 3, four days before its issuer was revoked.) eMudhra says every holder has been moved to a replacement.

Then there’s the question of how early eMudhra could have caught this. In June 2025, the certificate revocation lists for these intermediates stopped verifying. The signatures came from the original strong keys, while the certificates on file described the weak copies. eMudhra quieted the alarmOpens a new window by re-signing with whichever key matched each certificate, which meant the broken ones. It appears nobody asked why two sets of keys existed at all. The answer could have unraveled the whole thing more than a year early. Instead, the flaw sat in plain view for roughly 16 months.

Browsers were the easy part, again

Chrome, Firefox, Edge, and Safari cut off the compromised intermediates within days of the October disclosure. That’s good, but it still leaves everything in your environment that skips revocation checks. To drive the point home, a Czech developer has come up with a memorable demonstration of how much of it likely lingers on networks like yours.

Daniel Vrátil noticed his 2022 Samsung smart TV still trusted the compromised certificates, so he minted a certificate for Samsung’s ad domains with one of the leaked keys. Then he routed the TV’s traffic through mitmproxy and swapped the home screen adsOpens a new window for his family photos. (The tracking requests went into a black hole.) He published the project on GitHub, so you can see the whole experiment for yourself. It only works while the TV keeps trusting the dead CA. Vrátil warned of “countless much more sinister applications of this vulnerability.”

If you mentally swap the TV for anything in your rack that holds a private trust store, you’ll quickly understand why he issued this warning. Electron apps, Slack and Discord among them, bundle their own CA lists and only pick up the fix once you’ve updated the app itself. TLS-inspecting proxies vouch for whatever their own trust store accepts, so one stale appliance can quietly re-expose every patched browser you’ve got behind it.

Dennis Jackson, who dug into the keys on Mozilla’s behalf, publishes a tracker siteOpens a new window with test endpoints. They let you check your curl, your git, and any other client you’re unsure about with a command or two. Pay special attention to your out-of-support gear (like a Gen 6 firewall still humming in a closet after SonicWall’s support cutoff), since it will never get the memo at all. There’s no update coming for those boxes, so your options are segmentation now or retirement soon.

Although you’re probably not thrilled to have more cleanup duty back on your plate, the silver lining (if you can call it that) is that the remaining work comes with firm dates you can put on the calendar.

Already shipped. Apple’s macOS 27.2.1, 26.7.3, and 15.8.3 updates keep the bundled copies of curl and git from trusting the compromised CAs. They also drop five of eMudhra’s emSign roots from /etc/ssl/cert.pem. If your scripts start failing certificate checks on a freshly updated Mac, look at what the endpoint chains to before you blame your code.

By November 16. Microsoft starts pulling the last 10 eMudhra rootsOpens a new window from Windows, Edge, Azure Linux, and the rest of its product line. If you manage Windows machines that don’t get automatic trust updates, you’ll need to deliver those changes by hand. eMudhra’s hard stop on all issuanceOpens a new window from its final hierarchy, G1, lands the same day, and Chrome and Firefox won’t trust anything G1 issues after that point.

By November 30. Apple goes further and ends trust in every G1 certificate, including ones issued before the deadline, so don’t count on an older G1 certificate buying you time. eMudhra also owes a complete inventory of its CA keys that day.

December 4. KPMG, the independent reviewer eMudhra brought in, is expected to deliver its findings, which should tell you how much of this post-mortem will hold up.

A trust store removal behaves like a patch in reverse. The stragglers face the same race that plays out after any fix ships, in which attackers hunt whichever systems haven’t caught up yet. Every device of yours that misses these dates keeps trusting a hierarchy the major root programs have written off.

Audit your certificates like the vendor was wrong

Your October audit ran on assurances from eMudhra, but the company has now admitted those claims were premature. So, re-run it assuming the vendor’s all-clear was never accurate. You’ll want to search your inventory for certificates chaining to any emSign root, not only the distrusted G3 hierarchy. For your proxies and your odd embedded clients, let the tracker’s test hosts do the checking. If you can’t patch something, consider treating it as permanently exposed and segmenting it away from anything you care about. When this second cleanup finally closes out, you’ll know your exposure ended because you verified it yourself—not because the vendor said so.

I reached out to eMudhra for comment on this story, but the company did not provide one by publication time. I’ll update this reporting if that changes.