Caremark After Marchand: What It Means for the Family Council

The In re Caremark decision was a 1996 ruling by the Delaware Court of Chancery. Marchand v. Barnhill was a 2019 ruling by the Delaware Supreme Court. Between them, in the twenty-three intervening years, sat what most family-office and private-trust governance professionals treated as a public-company problem: the doctrine of board oversight liability for failures of monitoring around mission-critical risks.

That treatment is no longer defensible. In the two years between 2024 and early 2026, three things converged. First, Delaware courts began treating cybersecurity not as one risk among many but as a mission-critical risk subject to the heightened oversight standard articulated in Marchand. Second, the SEC's amended Regulation S-P imposed direct fiduciary-style obligations on registered investment advisers regardless of size — and most family-office adjacent advisers fall under that scope. Third, the operating reality of AI agents inside the family office stack created a new vector of board-level exposure that the 1996 doctrine did not contemplate but the 2026 doctrine plainly does.

What Marchand changed

The Marchand v. Barnhill decision arose out of the Blue Bell Creameries listeria outbreak. The Delaware Supreme Court held that the board of directors had failed to make a good faith effort to implement a board-level system for monitoring food safety — a risk that, for an ice cream company, was mission-critical. The board's records contained no committee dedicated to food safety, no regular reporting on food safety, no documented oversight discussion of food safety. The court ruled the directors had breached their duty of loyalty by failing to attempt to fulfill that monitoring obligation.

Mission-critical, in Marchand's language, is not "what the company spends the most money on" or "what shows up in the strategic plan." It is the risk whose realization would cause the firm to fail in its essential purpose. For Blue Bell, that was food safety. For a bank, that is the integrity of customer accounts. For a family office, that is the protection of the family's wealth and the family's privacy.

By 2025, Delaware courts had begun treating cybersecurity in exactly that frame. The Caremark line, after Marchand, now contemplates a board's affirmative obligation to implement a board-level system for monitoring mission-critical cyber risk: regular reporting, dedicated committee or sub-committee attention, documented discussion, a clear standard of escalation, an articulated risk appetite. Failure to attempt this exposes directors to personal liability for breach of the duty of loyalty.

The doctrine was written for public companies. The structure does not stop at the public-company boundary.

Why this reaches private trust companies and family councils

A private trust company exists, in most cases, to hold and manage family wealth across generations. The directors of that trust company — frequently family members, frequently augmented by independent professional directors — owe duties of loyalty and care under state trust and corporate law. Those duties are usually understood in the trust-administration frame: prudent investment, faithful execution of trust terms, proper accounting.

The 2025 reading of the Caremark line extends that frame. If the realization of a cyber risk would cause the trust to fail in its essential purpose — protection of family assets, protection of family privacy, integrity of the trust's records — then the directors have an affirmative obligation to implement a board-level monitoring system for that risk. The fact that the trust company is private rather than public does not change the analysis. Delaware fiduciary duties travel with the entity, not with whether the entity has public shareholders.

The family council sits in an adjacent but related position. Most family-council charters are not legally binding the way a corporate board's charter is, but the council members often serve simultaneously as directors of the family's private trust company, as trustees of one or more family trusts, as officers of family-owned operating businesses, and as members of investment committees with discretionary authority. In each of those capacities, fiduciary obligations apply. The council member who has cybersecurity within their portfolio of oversight responsibilities, however informally that portfolio is described, has an affirmative duty to monitor.

What the monitoring system actually requires

The minimum is a documented system. Not a perfect program. A system reasonably designed to surface, escalate, and address mission-critical cyber risk before its realization. The Marchand court was not asking whether Blue Bell had perfect food safety. It was asking whether the board had any system at all. The 2025 Delaware reading does not ask whether the family-office board has perfect cybersecurity. It asks whether the board has any system at all.

A reasonably-designed system has five components, all of which must be documented. A standing item on the board or council agenda for cyber and AI risk, with clear minutes capturing what was discussed and what was decided. A defined cadence of reporting — typically quarterly — from the executive responsible for cyber and AI to the board or council. A documented risk appetite statement that identifies the cyber and AI risks the entity considers mission-critical and the threshold at which a specific risk must be escalated. A clear escalation path that defines who, in what timeframe, must inform the board of an incident or near-incident. A documented framework — usually one of NIST Cybersecurity Framework, NIST AI Risk Management Framework, ISO 27001, ISO 42001, or a written internal hybrid — that the board has reviewed and approved as the operating baseline for the firm.

None of those five components requires technical expertise on the part of the board members. All of them require that the board members exercise judgment about the system, document that judgment, and demonstrate the ongoing application of that judgment over time.

The exposure that the 1996 doctrine did not contemplate

The exposure is asymmetric. Boards that have a documented system, even an imperfect one, are well-positioned under the Caremark line. Boards that have no system at all are exposed. The exposure is to personal liability under the duty of loyalty, not just to corporate damages — which is a category of risk that most family-office directors have not previously had to think about.

The exposure is also new in a specific way. The 1996 Caremark doctrine contemplated boards monitoring external risk: regulatory exposure, financial reporting failures, environmental and health and safety failures. The 2026 reading must contemplate internal risk created by the firm's own AI agents. An AI coding agent operating with credentials it should not have had can take destructive action with no malicious actor in the loop. An autonomous CIO research workflow can leak material non-public information to a third-party model provider. A contract-review agent can execute on a misread of an ambiguous clause and create downstream legal exposure. The Replit, Amazon Kiro, Cursor/PocketOS, and Amazon Q incidents documented across the last twelve months are the publicly available evidence that this class of incident is now real, recurring, and capable of producing seven- and eight-figure damages.

For a private trust company director who has not asked, in writing, whether the family office has runtime governance over its own AI agents — the answer to that question is now part of the Caremark-style monitoring obligation.

What to ask this quarter

Three questions belong on the next board or council agenda. What mission-critical cyber and AI risks does the family office face, and where are they documented? What is the current monitoring system for those risks, and when did the board or council last review it? What incidents — including near-misses — have occurred in the prior quarter, who escalated them, and how were they resolved?

The Marchand standard is not whether the board got every answer right. The standard is whether the board asked the questions, documented the answers, and exercised judgment over time. The directors who can demonstrate that posture on June 4, 2026, and in the years that follow, are operating inside the doctrine. The directors who cannot are exposed by it.

Brad McEvilly writes on agentic AI governance, fiduciary cyber duty, and the operating reality of family-office security. He is the founder of DeepSweep and the author of The Governance Gap (May 2026).