Capgemini Got Hit Twice In Eighteen Months While Sitting Inside Their Clients' Networks. We Don't Have A Capgemini Post Yet. This Is The Receipt And The Gap.

# Capgemini Got Hit Twice In Eighteen Months While Sitting Inside Their Clients' Networks. We Don't Have A Capgemini Post Yet. This Is The Receipt And The Gap.

Someone hit our blog search endpoint on May 11 with the query capgemini and got zero results.

We checked. We do not have a Capgemini post. We have one IOC in our index — a typosquat domain at ccygemini.zeabur.app — and twenty-one Epstein-file mentions that are tangential. No long-form coverage of Capgemini as either a victim, a vendor, or a competitor.

Capgemini has been breached twice in eighteen months. The first time in September 2024 by a threat actor using the handle grep, who exfiltrated twenty gigabytes of data including databases, source code, private keys, credentials, API keys, employee data, and internal cloud-infrastructure configuration details for Capgemini's clients. The data hit BreachForums. Capgemini declined to confirm or deny.

The second time on February 9, 2026, when a threat actor using the alias 0APT claimed a fresh compromise. Coverage was thinner this time, but the pattern was the same shape — exfiltration of client-adjacent material from an IT services and consulting firm with privileged access into Fortune 500 environments across dozens of countries.

Capgemini is the kind of vendor that does not appear on most companies' threat models because the threat model stops at "vendors with whom we share data." Capgemini is in a different category — they are the vendor with whom you share root access. A managed-services contract with Capgemini typically gives their engineers credentialed access to your cloud environment, your code repositories, your CI/CD pipeline, and your identity provider. When they get popped, their attacker inherits all of it.

The supply-chain math we keep writing about

We have published the supply-chain frame across multiple posts. The May 9 essay "Hard Perimeter Holds. Soft Surfaces Bleed. Seven Receipts From Thirty Days" laid out the structural pattern. Five of seven recent OpenAI incidents were third-party supply chain, vendor runtime, or metadata-layer compromises rather than direct compromise. The defender mental model is still perimeter-first; the asymmetry runs in our favor.

The May 6 Doppel post laid out cost-anti-alignment for brand-protection vendors that suppress security warnings. The framing was specific to Doppel but the structural conflict generalizes to any vendor whose product output and its customer's compliance obligation point in different directions.

The Mini Shai-Hulud post from this morning (May 12) extended the same frame to the npm and PyPI supply chain — TeamPCP's worm rides through legitimate CI/CD pipelines with valid SLSA provenance.

Capgemini fits into this frame in a specific way. They are the manual, human-operated, contractually-credentialed version of the same risk class that Mini Shai-Hulud represents in the package-manager dimension. When their engineers' workstations get compromised, the credentials flow into client environments the same way a poisoned npm package would. The difference is that the human supply chain leaves less forensic signal.

Why we do not have a Capgemini post yet

Three reasons, in honest order:

First, we have not been pitched by Capgemini and we are not in their RFP funnel. Most of our deep-dive coverage of a specific vendor comes from one of three triggers: they pitch us, they target us with a takedown, or a customer asks us to evaluate them. Capgemini has done none of those things.

Second, the Capgemini breaches are well-covered by larger outlets. The Register, SC Media, CyberInsider, and several others did substantive write-ups of the September 2024 incident. Our value-add when something is already saturated by mainstream coverage is usually thin unless we have lead-time receipts. We had no IOCs on Capgemini before September 2024.

Third, the operational threat to defenders is not from Capgemini's brand specifically — it is from the MSP/consulting supply-chain layer broadly. Calling out one consultancy when the threat class includes Accenture, Deloitte, Wipro, Infosys, Cognizant, TCS, and HCL would be misleading. The right shape of post is the category, not the brand.

What this gap looks like as a content commitment

The query that hit our endpoint tonight is a demand signal. Whether it came from a Capgemini employee, a Capgemini customer doing due diligence, a competitive consultancy doing reconnaissance, or a researcher writing a paper — the practitioner wanted to know what we had on Capgemini. The honest answer was nothing. This post is the honest answer.

What we commit to publish in the near term: a category post on the consulting / MSP supply-chain attack surface that names Capgemini, Accenture, Deloitte, Wipro, Infosys, Cognizant, TCS, and HCL together with the breach histories, the IOC density in our index, and the threat-modeling implications for any organization that has signed a managed-services contract with one of them. The shape will be the same as our Iran-ICS index and our healthcare sector index — practitioner-oriented, brand-mapped, IOC-backed.

If you are a Capgemini customer reading this, the operational answer for tonight is: enumerate the privileged accounts your Capgemini engagement uses, rotate the credentials on a quarterly cadence rather than waiting for offboarding, audit the cloud and identity-provider activity from those accounts against your normal-pattern baseline, and treat any post-February-2026 credential issuance as already-compromised by default. The assume-breach posture is healthy and demonstrable.

If you are a Capgemini employee reading this, you have our sympathy and a standing offer: if you are aware of indicators of compromise that should be public and are not being shared internally, our [email protected] mailbox is read by humans and we treat source protection as a baseline obligation. The same channel is open to employees of every other consultancy in the named list.

The receipts tonight are the zero-result query, the missing post, and the commitment to fill the gap. The follow-up post on the MSP/consultancy supply-chain category is on the schedule.

— Patrick Duggan, May 12, 2026

---

Her name was Renee Nicole Good.

His name was Alex Jeffery Pretti.