Can HOA Boards Use AI Without Waiving Privilege?
Consumer chatbots turn board decisions into discoverable records and can waive privilege the moment someone pastes a lawyer's memo. A closed community-trained agent changes the math.
The short answer
HOA boards can use AI without waiving attorney-client privilege only if inputs stay inside a closed data boundary the association controls. Pasting a privileged attorney memo into a public chatbot like ChatGPT can waive privilege on the whole file, because sharing it with a third party breaks confidentiality and can make every prompt discoverable in litigation.
The deposition question nobody prepped the board for
"Mr. President, did the board use ChatGPT or any AI tool to prepare its response to the plaintiff's demand letter?" Opposing counsel asks it flat, on the record, in a construction defect deposition. The board president pauses. A director actually did paste the attorney's memo into a chatbot to "tighten up" the owner communication. Nobody told the lawyer.
That pause is the whole problem. Once the answer is yes, the next demand is predictable: produce every prompt, input, and output tied to this dispute. And because the memo was shared with a third-party system outside the attorney-client relationship, plaintiff's counsel now argues privilege was waived, not just on that memo but on the related file.
This is not a hypothetical about robots taking jobs. It is a records and evidence problem, and it is already reaching community associations that treated consumer AI as a harmless shortcut.
The three sentences that should scare every board
Key takeaways
- Anything typed into a public chatbot may be a discoverable record: a subpoena can demand your prompts, uploads, and outputs.
- Pasting a privileged attorney communication into a third-party tool can waive privilege, because confidentiality is the thing privilege protects.
- A waiver rarely stays contained: courts can extend it to the whole subject matter, dragging in memos the board never meant to share.
Attorney-client privilege is fragile by design. Privilege is the legal protection that keeps confidential communications between an association and its lawyer out of opposing hands. The instant that communication is voluntarily shared with an outside party, the confidentiality it depends on can evaporate.
A public chatbot is an outside party. When a director pastes counsel's litigation strategy into a consumer tool, that is arguably a voluntary disclosure to a third-party service, terms of use and all. The board's intent to stay private does not automatically save the privilege.
Why public AI tools create discovery exposure
Direct answer
Public AI tools create discovery exposure because your prompts and uploads become records held by a third party, and records can be subpoenaed. When those records contain legal advice, the act of sharing them outside the attorney-client relationship can waive privilege and open related documents to disclosure.
In litigation, discovery reaches anything relevant that a party controls or can obtain. If a board member used a consumer chatbot to draft a violation response, an assessment notice, or a reply to a demand letter, those interactions are potentially discoverable. Opposing counsel does not need proof first: they can ask, in writing and under oath, whether AI was used at all.
The FBI's Internet Crime Complaint Center has spent years documenting how casually shared data becomes leverage against the people who shared it. The lesson for boards is the same: information that leaves your control stops being yours to protect.
The uncomfortable part: most boards have no log of who typed what into which tool. There is no audit trail, no retention policy, and no way to answer the deposition question honestly because nobody knows the full answer.
Consumer chatbot vs a closed community-trained agent
The difference that matters legally is the data boundary: where your inputs live, who can reach them, and whether they train a model the rest of the world uses. A closed community-trained agent keeps inputs inside the association's walls and never feeds a public model.
| Factor | Public chatbot (ChatGPT, free tier) | Closed community-trained agent |
|---|---|---|
| Where inputs go | Third-party servers under consumer terms | Inside the association's controlled boundary |
| Trains a public model | Often yes on consumer tiers | No, scoped to your community only |
| Discovery footprint | Prompts may be discoverable third-party records | Records stay under association control and policy |
| Privilege risk | Sharing counsel's memo may waive privilege | No third-party disclosure of the memo |
| Audit trail | Usually none the board controls | Logged, attributable, retention-governed |
| Who sees the data | The vendor and its systems | The community, its manager, and its counsel |
None of this makes the closed agent a lawyer. It makes it a safe drafting surface. The privilege still lives with counsel; the agent just stops you from broadcasting privileged material to an outside service by accident.
The SAFE-PASTE checklist for boards and CAMs
Use this before anyone types into any AI tool. The rule of thumb: if a passage could be evidence in a dispute, it does not go into a public chatbot, ever.
Checklist
0/10What may and may never enter a public AI tool
Where a closed agent keeps privileged work inside the walls
This is exactly the gap One Home Agent built its property management agents to close. Bailey Board assembles board packets, minutes, and action items using only that community's own records, inside a boundary the association controls, so sensitive material never crosses into a public model.
CAMeron, the community manager copilot, holds institutional memory per community: prior counsel guidance, past board decisions, the reasoning behind them. When a manager needs to draft a response consistent with what the attorney already advised, the agent works from documents that stayed private the whole time.
The point is not that the agent is smarter than your lawyer. It is that the drafting happens where the records already live, so no director has to paste a privileged memo into a consumer app to get a first draft. The boundary is the feature.
“The single biggest legal reason to stop using consumer AI in an HOA is not accuracy, it is confidentiality. The moment a privileged memo lands in a public chatbot, the board has handed opposing counsel an argument it did not have to give. Keep the drafting inside the boundary and let the attorney keep the judgment.”
Todd Paton, Partner, One Home Agent
AI absorbs the drafting, your attorney keeps the judgment
The right division of labor is simple. The agent absorbs the repetitive, documented, deadline-driven drafting: assembling the packet, pulling the relevant history, producing a first pass that matches prior guidance. Your attorney keeps the judgment: strategy, privilege calls, and the final word on anything headed toward a dispute.
Bottom line
Boards can use AI safely, but not the consumer kind on privileged material. Keep every sensitive input inside a closed, community-trained boundary, keep a written policy and log, and keep your attorney in the approval seat. Do that, and AI cuts the busywork without ever handing opposing counsel a waiver argument.
Give your communities an agent that keeps privileged work inside the walls
See how a community-trained agent handles board packets, minutes, and manager drafting without exposing privileged material to a public model. The first agent we build for your company is free, and you keep it.
See the PM agentsFrequently asked questions
It can. Privilege depends on confidentiality, and pasting a privileged memo into a public chatbot arguably shares it with a third-party service outside the attorney-client relationship. Courts can find that voluntary disclosure waived privilege, sometimes extending the waiver to related documents on the same subject.
Sources & further reading