Privacy Computation Investment Counsel — Health Data Alliance

The mandate

The supervisory board repeatedly faces the same investment question: which combinations of federated analysis, trusted execution, multiparty computation, differential privacy and synthetic data can unlock cross-institution research without making promises that fail under clinical or adversarial scrutiny? Member proposals use similar privacy language despite radically different leakage, performance, governance and scientific-validity trade-offs.

The adviser will contribute three days each month: one chair and committee session, one technical challenge clinic with project teams, and one day for paper review or member consultation. Five scheduled Research Governance Committee meetings are included. Advice on a suspected disclosure or material partnership term must be returned within two Dutch business days, with extraordinary incident work separately agreed.

The appointment runs for ten months beginning January 2027. In month eight, the supervisory board will determine whether the alliance has built enough internal judgement to close the role; it may approve one six-month renewal by recorded vote. Neither management nor an individual member institution can extend the retainer unilaterally.

This is an influence-only appointment carrying no line authority, executive responsibility or power to approve studies, process patient data, select vendors or bind member hospitals. Accountable executives and ethics bodies retain every decision. Advice will identify assumptions and dissent so the record cannot be misread as delegated regulatory or clinical approval.

The adviser may retain up to two unrelated appointments, but must disclose work for privacy-technology vendors, competing data alliances, participating pharmaceutical sponsors or member institutions. A paid role with a supplier under active selection requires recusal; access to overlapping study protocols or datasets may require the board to terminate one engagement rather than rely on information barriers.

Why the board wants this voice

The room contains distinguished clinicians, privacy lawyers and research executives, yet nobody has taken privacy-preserving computation from laboratory claim to repeatable multi-party operation. Directors need a voice able to challenge both mathematical assurance and practical data leakage without becoming another vendor advocate. The intended legacy is better board judgement, not a permanent technical oracle.

What you will own

• Press directors to distinguish policy confidentiality, cryptographic protection, statistical disclosure control and contractual trust when proposals blur them together.
• Test whether each proposed technique preserves the scientific utility, cohort representativeness and reproducibility needed for its stated clinical question.
• Challenge threat models for collusion, auxiliary information, repeated queries, malicious participants, compromised enclaves and model inversion using plausible operating conditions.
• Shape investment gates that connect privacy claims to measurable leakage budgets, performance limits, key management, auditability and accountable residual-risk acceptance.
• Probe commercial dependencies created by proprietary clean rooms, specialist hardware, orchestration layers and cross-border support arrangements before commitment.
• Frame member-alliance principles for publishing negative findings, responding to disclosure, withdrawing a dataset and communicating limitations to research participants.
• Coach the committee to compare pilot evidence across techniques without treating incompatible metrics or vendor demonstrations as an objective portfolio ranking.

Candidate qualifications

• Governed privacy-enhancing computation in a production health, life-sciences, financial or public-data collaboration involving legally independent data controllers.
• Can interrogate the assumptions behind multiparty computation, differential privacy, federated learning and trusted execution without claiming every method fits every workload.
• Challenged a privacy claim using empirical leakage or adversarial testing and influenced a material design, investment or study-governance decision.
• Worked with ethics, clinical, legal and security leaders where scientific usefulness and disclosure protection could not both be maximised.
• Advised a board or consortium council on technology whose residual risk remained partly uncertain and ensured that uncertainty was recorded honestly.
• Managed portfolio conflicts involving vendors, research sponsors and institutions with overlapping datasets, intellectual property or competitive interests.

Non-negotiables

• Able to attend all five committee meetings in Amsterdam and preserve the three-day monthly commitment for the full term.
• Will disclose investments, research funding, expert-network participation and advisory relationships touching any relevant technology or member institution.
• Accepts that the board may reject advice and that accountable ethics, privacy and executive officers retain decision and legal responsibility.
• Brings deployed-system evidence; academic publication or privacy-policy leadership without operating exposure does not meet the mandate.

1. 49 words maximum. Describe a privacy-enhancing computation claim you challenged and the evidence that changed the investment or deployment decision.
2. 49 words maximum. Which current commitments could intersect with alliance members, sponsors or privacy-technology suppliers, and how would you manage them?
3. 49 words maximum. How would you test whether a differential-privacy budget preserves both disclosure protection and a study’s scientific utility?