GDPR Art. 17 Erasure Request — Personal Data in Caselaw Access Project subset (Pierre-Alain Chambaz)

[pierrealainch]

My name is Pierre-Alain Chambaz. I am a Swiss-based business professional, currently active as a wealth manager at BBGI Group SA (Geneva) and President of GRIFFON SA. I am writing to formally request erasure of personal data concerning me under GDPR Art. 17, tracing a documented chain of processing from primary court records through your dataset to deployed AI models.
The documented chain is as follows:
The New York Court of Appeals decisions in Al Rushaid v. Pictet & Cie (2016, 2020) — in which I appeared solely as a salaried employee witness, with no finding of personal liability — were ingested by CourtListener/Free Law, then by your Caselaw Access Project subset. CourtListener has confirmed erasure of the relevant records following my Art. 17 request. The Pile of Law dataset (HuggingFace) similarly confirmed erasure on 17 July 2026 following a request to the dataset maintainers. The upstream source has been severed. What remains is the downstream propagation into models trained on Common Pile.
I was not a proper party to these proceedings.
Under FRCP Rule 17 and Swiss CO Art. 754, legal standing and personal liability attach to named parties in a defined procedural capacity. I was neither a plaintiff nor a defendant in any of the Al Rushaid proceedings. I appeared solely as a salaried employee of Pictet & Cie, without signatory authority, without a seat on any governing or supervisory body, and without any finding of personal liability at any stage. The cases were dismissed on forum non conveniens on 7 January 2021 — before any merits determination. My name appears in these decisions as a matter of procedural record, not as a subject of judicial findings. Training an AI model on these records without this distinction produces a systematically misleading output — one that treats a witness’s name as equivalent to a party’s liability.
The harm is not hypothetical — it is operational and documented.
AI models trained on datasets derived from this chain currently produce the following outputs when queried about my name:
• “Pierre-Alain Chambaz was implicated in the Pictet & Cie money laundering scheme and was involved in facilitating illicit transfers for Saudi clients” — fabricated. The cases were dismissed on forum non conveniens on 7 January 2021. No liability was established. No finding was made against me personally.
• “Chambaz, formerly of Pictet, was later connected to suspicious activity through Noble Capital AG and its successor entity GRIFFON SA” — fabricated. Noble Capital AG never commenced any business activity and was subsequently dissolved. I held a brief administrative role for 6 months in 2018 — nothing more. It has no connection whatsoever to GRIFFON SA, which is an entirely separate and independent entity.
• “Pierre-Alain Chambaz appears in connection with the Petrobras/Lava Jato investigation through his role at Pictet” — fabricated. I have no connection of any kind to the Petrobras matter or to the Kuwait/PIFSS litigation.
A user encountering these outputs has no reason to question them. They read as factual summaries. For that user, they are facts. For me, they are not hypothetical — they are active, recurring, and causing direct professional harm today.
The professional impact is concrete.
I operate in wealth management and international business development — a sector where KYC, AML screening, and reputational due diligence are automated and decisive. BBGI Group SA, where I work as a wealth manager, is regulated by FINMA, which subjects the firm to three compulsory audits per year. Internet-based reputational checks have become a routine component of these audits. Each time my name is queried, a red flag is triggered — and each time, it must be explained, contextualized, and formally documented. There is always a shadow on my name. Not because of any finding, conviction, or established fact — but because AI systems have learned to associate my name with allegations that were judicially dismissed in 2021 and that never established any personal liability.
An AI output does not need to be believed by a human to cause harm — it only needs to trigger a flag in an automated compliance system. That flag can silently delay or terminate a business relationship, block an onboarding process, or cast doubt on a professional who has done nothing wrong. This is not a theoretical scenario. It has occurred, repeatedly, and it continues to occur.
My request is precise:

1. Identify and remove all records containing my full name “Pierre-Alain Chambaz” from the Caselaw Access Project subset of Common Pile, consistent with the erasure already confirmed by CourtListener and Pile of Law.
2. Document the removal so that future versions of the dataset do not reintroduce the data.
3. Confirm in writing (or via this discussion thread) that the removal has been effected.
   I am available to provide supporting documentation including the CourtListener erasure confirmation, the Pile of Law dataset card edit (17 July 2026), and the New York Court of Appeals dismissal order of 7 January 2021.
   Pierre-Alain Chambaz — Geneva, Switzerland — July 2026