Directors are already using AI for board work. That should prompt boards to think carefully about how these tools handle sensitive information.

Not because directors should avoid AI. Quite the opposite: AI may be one of the most valuable analytical tools available to a director. The concern is what happens when a director takes something that would ordinarily never leave the boardroom and enters it into a general-purpose chatbot.

A confidential concern about the CEO, an unpublished forecast, a potential acquisition, an activist vulnerability, a disagreement with management, or an entire board book. What feels like a private exchange with AI may create a record of sensitive boardroom thinking in a system outside the organization's control. That is a risk worth understanding.

The Question You Would Never Put in an Email

Directors already understand this principle in another context. There are things you might appropriately raise in a board meeting that you would be extremely reluctant to put casually into an email.

The reason is straightforward. Email creates a record. The words may persist long after the conversation is over. If litigation, an investigation, or another legal process arises, relevant electronic records may become subject to preservation and production requirements.

Now consider some of the questions directors might ask AI:

• Is our CEO the right person to lead the company through this transition?

• What weaknesses in our strategy could an activist exploit?

• Management wants us to acquire this company. Here are the board materials. Find every reason we shouldn't.

• These are our unpublished forecasts. Where is management being too optimistic?

These are not inappropriate questions. They are exactly the kinds of questions directors should be exploring. But imagine seeing the prompts, responses, and supporting documents reproduced years later, separated from the context in which a director was simply testing an idea.

AI makes it remarkably easy to create a detailed written record of questions that previously might have existed only as part of confidential deliberation. And the more useful AI becomes, the more sensitive that record may be.

Then Come the Documents

Typing a sensitive question is one thing. Uploading the underlying material raises the stakes considerably.

A director seeking a useful answer may naturally provide AI with a board book, financial model, acquisition materials, strategic plan, management presentation, or other internal documents. From the user's perspective, this makes sense: provide more context and receive a more informed answer.

From an information-governance perspective, however, something significant has happened. Highly sensitive corporate information has been introduced into another technology environment, potentially subject to different policies governing access, storage, and retention.

That distinction can easily disappear behind an interface that feels as casual as sending a text message. For board-level information, it deserves careful consideration.

"We Don't Train on Your Data" Does Not Answer Every Question

AI providers have made substantial progress on privacy. Business and enterprise products can offer meaningful protections, including restrictions on model training, temporary conversations, and different retention arrangements.

But boards need to ask a more precise set of questions. Is the conversation used for model training? Is it retained, and for how long? Where is the information processed and stored? Who can access it? What happens to uploaded documents? Are there circumstances in which information must be preserved for legal or security purposes?

These are separate considerations, and understanding the differences matters.

Some consumer AI products may use conversations to improve future models, depending on the product and the user's settings. That does not mean another user can simply retrieve someone's private conversation. Leading AI providers employ safeguards intended to protect sensitive information, although research has demonstrated that language models can sometimes memorize and reproduce portions of training data.

For directors considering whether to enter confidential corporate information into an AI system, understanding whether that information may be used for model training is therefore important. But opting out of training does not necessarily address retention.

A conversation may be excluded from model training and still remain on a provider's systems for some period of time. Similarly, a temporary conversation is not necessarily one that leaves no retained records.

For ordinary AI use, conventional retention arrangements may be entirely appropriate. Asking AI to improve a presentation or explain an industry concept is one thing. Using it to examine confidential board deliberations is another. We believe those activities warrant different levels of consideration when it comes to information privacy.

Boards Shouldn't Have to Choose Between AI and Confidentiality

Directors have compelling reasons to use AI. Their responsibilities include challenging management, evaluating competing perspectives, identifying risks, and understanding complex developments without managing the company's day-to-day operations. AI can be exceptionally valuable in supporting that work.

A director should be able to ask: What am I missing? What would make this strategy fail? Does the external evidence support what management is telling us? What is our strongest competitor doing differently? What assumption in this board book should I challenge?

One approach to managing information risk would be to prohibit directors from entering sensitive board materials into AI systems. While restrictive policies may be appropriate for certain tools or circumstances, they can also limit directors' ability to benefit from powerful analytical capabilities.

We built AI Analyst because we believe boards should have access to advanced AI without unnecessarily compromising the confidentiality of their work.

What If There Were No Conversation History to Retrieve?

AI Analyst was designed with a different approach to sensitive information because the boardroom presents different requirements.

Privacy and security were central considerations from the beginning of development. Rather than treating sensitive conversations as information to be routinely stored and protected afterward, AI Analyst was designed around eliminating unnecessary retention.

Our approach is to avoid maintaining a permanent content history of completed AI Analyst conversations. Prompts, responses, and session materials are treated differently from the operational information required to maintain a secure and reliable platform. Customer content is also excluded from AI model training.

This distinction is fundamental. The objective is not simply to restrict access to a stored archive of sensitive conversations, but to eliminate the need for that archive in the first place.

A director exploring an acquisition, questioning management's assumptions, or investigating a potential governance concern should not have to create an unnecessary permanent record of that analysis simply because AI was involved.

Privacy was not something we added to AI Analyst after building the product. It was one of the principles that shaped its development from the outset.

Privacy Is Only Half of AI Analyst

There is another limitation to using a general-purpose chatbot for board work: it does not have direct access to the breadth of intelligence BoardRadar brings together.

AI Analyst provides the conversational capabilities directors expect from a leading large language model. It also brings together BoardRadar's intelligence on companies and their competitors, including financial performance, workforce and talent, competitive activity, governance and risk, reputation, AI developments, market conditions, and other continuously tracked signals.

That changes the nature of the analysis.

Ask a general-purpose chatbot why a competitor is becoming more efficient, and it can reason through possible explanations based on the information available to it. With AI Analyst, directors can explore the question using relevant BoardRadar intelligence, including changes in revenue per employee, workforce movements, leadership changes, partnerships, AI initiatives, financial performance, and competitive developments.

Directors can also bring their own information into the analysis. Board materials and internal company documents can be considered alongside BoardRadar's outside-in intelligence, allowing directors to question assumptions, investigate emerging risks, and evaluate strategic decisions with broader context.

The question is no longer simply, What could explain this? It becomes, What does the available evidence suggest is actually happening?

That combination is what distinguishes AI Analyst. It brings together the reasoning capabilities of advanced AI, BoardRadar's continuously updated intelligence, and the information directors choose to provide, all within an environment designed with the sensitivity of boardroom work in mind.

Boards Should Be Able to Ask the Questions That Matter

Effective governance depends on directors being willing to ask difficult questions. Some will be uncomfortable. Some will challenge management. Others will explore possibilities that never materialize or assumptions that ultimately prove incorrect. This is a necessary part of thoughtful oversight and informed decision-making.

AI can make that process significantly more powerful by helping directors examine information, challenge assumptions, and consider perspectives they might otherwise overlook. But realizing those benefits requires confidence in how sensitive information is handled.

At BoardRadar, we believe AI in the boardroom should provide directors with greater analytical capability, better information, and more freedom to investigate what matters. It should allow them to combine the reasoning power of leading AI models with intelligence about what is happening inside and outside their company, without creating unnecessary permanent records of their deliberations.

That belief shaped the development of AI Analyst and continues to guide our approach to privacy and security.

Directors should be able to explore the questions that matter most, with technology designed to respect the confidentiality of the process.