董事会在人工智能治理中的角色

Artificial intelligence governance is not an exclusively technical topic. It is a matter of corporate strategy that demands direct attention from the board of directors.

According to MIT CISR research with 2,800 companies, only 26% of boards are digitally prepared to effectively oversee AI initiatives. The difference between these boards and the rest is not subtle: AI-savvy boards deliver return on equity 10.9 percentage points above average.

The Stanford HAI documented 233 AI incidents in 2024, a 56.4% increase from the previous year. These are cases of algorithmic bias, automated decisions without traceability, and privacy violations that generate regulatory, financial, and reputational impact.

The question is not whether the board should be involved with AI. It is whether it is prepared to do so with the depth the moment demands.

The NACD research reveals a paradox: 95% of organizations invest in artificial intelligence, but only 34% incorporate AI governance into their oversight structures. Fewer than 25% have AI policies approved by the board. Only 15% of boards receive regular metrics on intelligent system performance.

This means most boards approve AI budgets without visibility into what is being built, which decisions are being automated, and what level of operational and regulatory risk is involved.

The OORT AI Assessment exists precisely to fill this gap: it maps intelligent systems in operation, classifies risks, and delivers a clear overview to the board before scaling.

McKinsey confirms: only 39% of Fortune 100 companies disclose any form of AI oversight by the board. In only 17% of cases, artificial intelligence governance is an explicit responsibility of the board of directors.

The EU AI Act, in effect since August 2024, is the world's most comprehensive regulatory framework for artificial intelligence. Compliance obligations for high-risk systems take effect in August 2026. Fines can reach EUR 35 million or 7% of annual global revenue, whichever is higher.

In Brazil, Bill PL 2338/2023 was approved by the Senate in December 2024 and is being processed by the Chamber of Deputies. The bill creates the National AI Regulation System, coordinated by the ANPD, and requires algorithmic impact assessments for high-risk systems. Companies operating in the Brazilian market need to prepare for transparency and traceability requirements that follow the same European logic.

The Stanford HAI recorded 59 new AI regulations introduced in 2024, double the previous year. Gartner projects that by 2030, AI regulation will extend to 75% of global economies, creating a compliance market that will surpass US$ 1 billion.

Boards that lack visibility into which intelligent systems operate in their company are, in practice, assuming an invisible regulatory liability.

Boards that effectively oversee AI share a consistent structure. It is not about creating a bureaucratic layer over technology, but about ensuring the board has the right instruments to exercise its fiduciary function.

AI Inventory is the starting point. Many boards are unaware of how many intelligent systems operate within the organization. Without a complete mapping, there is no way to assess risk exposure. The first step is knowing what exists.

Risk classification determines which systems require more rigorous oversight. Systems that make decisions about people (hiring, credit, healthcare) demand different controls than systems that optimize logistics or recommend content. The EU AI Act defines four categories: unacceptable, high, limited, and minimal risk.

Oversight committee is the organizational structure that connects operations to the board. Deloitte identified that 22% of boards delegate AI governance to the audit committee and 25% to the risk committee. Companies like Microsoft operate with four governance layers: board committee, AETHER council, responsible AI council, and operational team.

Metrics and reporting ensure the board receives actionable information. Not technical reports on model accuracy, but impact indicators: costs avoided, incidents detected, compliance status, team adoption. Only 15% of boards receive this type of information today (McKinsey).

Auditing and traceability close the cycle. Every decision made by an intelligent system needs to be recorded, explainable, and reversible. This is not precaution. In jurisdictions like the EU, it is a legal requirement. The OORT platform incorporates traceability by design: every agent action is recorded, auditable, and reversible.

Most companies try to apply governance over AI systems that were not designed to be governable. The result is a bureaucratic layer that reduces speed without reducing risk.

OORT's approach is different. The AI-First data layer ensures that information is ingested, normalized, and governed before feeding any intelligent agent. Every data point has traceable origin, version, and access control.

The OORT Flows agents operate with complete action logging. Every automated decision can be audited, explained, and reversed. For a board of directors, this means having confidence that artificial intelligence operates within the limits defined by corporate governance.

And the OORT Culture program prepares leaders and teams to operate with AI responsibly, ensuring that adoption does not outpace the organizational capacity to oversee.

The risk of neglecting AI oversight is concrete: regulatory fines, bias incidents, non-auditable automated decisions, and loss of institutional trust. But the opportunity cost of not governing well is equally relevant.

Boards that structure AI governance are not slowing innovation. They are creating the conditions for it to occur sustainably, scalably, and defensibly before regulators, investors, and society.

MIT demonstrated that these boards deliver superior financial results. EY showed that the market is demanding transparency. Regulation is coming, both in Europe and Brazil. The time to structure oversight is now, not when the first incident occurs.