AI Usage: Why “Share” vs. “Don’t Share” Is a False Sense of Safety

AI discovery risk is a new governance problem for any organization using generative AI in day-to-day work. Courts have indicated that AI prompts and outputs can be treated as discoverable electronic records in litigation, on the same footing as emails and other electronically stored information, if they are relevant and not privileged. A simple “share or don’t share” rule does not address that.

Core Principle: Control Starts With the Question

The primary control is what employees ask AI to analyze. Before sending a prompt, employees should be able to answer three checks:

  • Is this conceptual, or does it include confidential facts?
  • Is this based on public information, or on client, employee, or customer data?
  • Does it touch strategy, pricing, legal exposure, or active disputes?

If the prompt involves confidential, privileged, or regulated data, it must not be sent to consumer AI tools. It should be used, if at all, only through approved channels and with legal guidance.

Managing AI Discovery Risk With Public AI Tools

For hosted AI, such as consumer chatbots, the organization should consider three rules.

  • Ban red-line content in prompts. That includes customer or employee personal information and health information, trade secrets (code, formulas, algorithms, pricing logic, proprietary models), and live M&A, investigations, disputes, or board-level strategy.
  • Limit use cases. Stay with market research, best-practice summaries, generic training, and policy drafting. Keep company-specific contracts, live deal decks, and internal financials out.
  • Require “strip and abstract.” Remove names, exact amounts, and unique identifiers. Ask for pattern-level guidance, such as “a services firm with high client concentration,” instead of pasting raw internal data.

AI Discovery Risk Does Not Disappear With Private AI

Running AI on local machines or in private environments does not remove discovery risk. It shifts it in-house. The organization should consider:

  • Treat AI artifacts as discoverable working papers. Assume prompts and outputs related to legal, financial, or strategic matters may be requested in discovery and preserved under legal holds.
  • Integrate AI into existing governance. Classify AI files (prompts, outputs, drafts) under the same record types as their source material, such as legal, HR, and finance. Apply the same retention, legal-hold, and access controls used for spreadsheets, emails, and memos.
  • Define approved local-only use. Permitted examples include cleaning up a template model, generic scenario generation, and non-confidential training materials. Higher-risk uses, such as full general ledger exports, email archives, or draft complaints, need prior approval from Legal and IT.

The Practical Takeaway for Leadership

Whether employees use public AI that learns from prompts or private AI on company devices, the economic reality is the same. They are creating new digital records that may be discoverable, hackable, and auditable. Leadership’s role is not to ban AI. It is to determine where the risk lies, set clear usage boundaries, and govern AI activity like any other source of electronic evidence. Managing AI discovery risk is a governance task, not a technology one.

This article is general business information, not legal advice. Discovery rules and court decisions vary. Consult your attorney about your specific situation.

Book a call with Business CFO for Hire to set AI usage boundaries that fit your company.

Share this:

About the Author

Stan Alhadeff, fractional CFO and author of Run the Business Don't Become It, featured in Atlanta Business Journal Leaders in Finance

Stan Alhadeff is the founder of Business CFO For Hire and one of the longest-serving independent fractional CFOs in the U.S. With 30+ years of financial and operational leadership spanning startups to $1B+ enterprises, he's guided companies through fundraises, ownership transitions, rapid growth, and M&A across a dozen-plus industries. He's also the author of Run the Business, Don't Become It, a field guide to financial clarity for founders and CEOs. Based in Atlanta, Stan works with growing businesses nationwide.

SIGN UP

Business CFO Insights Newsletter