Law firms in every US state and every Canadian province can ethically use an AI receptionist today. No bar or law society prohibits it, and the governing framework is not new: the duties of confidentiality, competence, and supervision that have always applied to human receptionists, answering services, and cloud software apply identically to AI intake. What changes is the diligence checklist — where call data lives, whether calls are recorded and disclosed, and how the system is fenced off from giving legal advice.
This article is the reference version of that checklist for both countries: privilege and confidentiality analysis, recording-consent law state by state and province by province, and the actual regulator guidance — ABA, state bars, Canadian law societies. For the implementation side (greetings, scripts, escalation), pair it with our AI receptionist setup guide.
No. In both the US and Canada, privilege extends to communications made through a lawyer's agents for the purpose of obtaining legal services. A prospective client's call answered by a firm's receptionist, answering service, or intake staff has always been treated as protected — in the US as attorney-client privilege plus the prospective-client duties of ABA Model Rule 1.18, in Canada as solicitor-client privilege, which the Supreme Court of Canada has repeatedly described as near-absolute. An AI receptionist acting as the firm's intake agent sits in the same legal position as the human it replaces.
The genuine risk is not waiver-by-robot; it is confidentiality in the vendor's hands. Privilege protects against compelled disclosure; it does not secure a database. If a vendor stores transcripts insecurely, uses your callers' facts to train shared models, or hands data to third parties, the firm has a confidentiality problem under ABA Model Rule 1.6 or the Canadian codes of conduct regardless of what a court would say about privilege. That moves the analysis to vendor diligence, covered below.
One nuance deserves emphasis for prospective clients specifically. Under ABA Model Rule 1.18 and the Canadian equivalents, duties of confidentiality attach to would-be clients even if the firm never takes the matter. That means the intake transcript of a caller you decline is still confidential information the firm must protect, and it means the intake script should collect enough to screen conflicts without vacuuming up more sensitive detail than the screening requires. A disciplined AI script is actually easier to keep within these bounds than a free-ranging human conversation — the AI asks exactly the questions it is configured to ask, every time.
Most AI receptionist platforms record calls or retain transcripts, so recording law applies from day one.
Two edge cases to configure deliberately. Cross-border calls: consent law generally follows the strictest jurisdiction touched by the call, so a firm anywhere in North America should simply run the all-party-consent greeting on every call rather than attempting geographic detection. And transcript-only setups: even where audio is not retained, a verbatim transcript is still a record of the communication — Canadian privacy regulators and cautious US counsel treat the notification obligation as applying either way.
The practical solution is one sentence in the greeting: "This call may be recorded and transcribed for quality and intake purposes." Spoken before any substantive exchange, it satisfies all-party consent states and Canadian notification requirements simultaneously. Configure it once; never think about the state-by-state map again.
The controlling framework is ABA Model Rule 1.6 (confidentiality), 5.3 (supervision of nonlawyer assistance — which the ABA has long read to include technology vendors), 1.1 comment 8 (technology competence, adopted in 40+ states), and 1.18 (prospective clients). On generative AI specifically, ABA Formal Opinion 512 (July 2024) requires lawyers to understand the AI tools they use, safeguard confidential information — including assessing whether inputs train the provider's models — supervise output, and communicate with clients where appropriate.
State-level guidance is consistent: the California bar's 2023 practical guidance on generative AI stresses confidentiality and vendor terms; Florida's Ethics Opinion 24-1 permits generative AI with confidentiality and oversight safeguards; New York, Texas, and others have issued reports in the same key. None of it prohibits AI intake. All of it converges on three operational duties: vet the vendor, disclose appropriately, and keep a lawyer supervising the system's configuration and output. A firm that reviews AI call transcripts on a set cadence and controls its scripts is doing precisely what Rule 5.3 asks.
Canada regulates lawyers provincially, and the Federation of Law Societies' Model Code — adopted with variations everywhere — contains the same building blocks: confidentiality (s. 3.3), supervision of non-lawyer staff and services (s. 6.1), and technological competence in the commentary to the competence rule. The Law Society of Ontario's guidance on generative AI (2024) and its practice-management guidance on outsourcing both apply cleanly to AI reception: understand the tool, protect confidentiality, supervise, and remain responsible for the work. The Law Society of British Columbia's guidance on AI strikes the same notes, and Alberta, and the Barreau du Québec have issued equivalent material.
Two Canada-specific overlays matter. Data residency: no federal rule forbids storing call data in the US, but PIPEDA requires comparable protection and transparency about cross-border processing, some public-sector-adjacent work carries stricter requirements, and Quebec's Law 25 requires a privacy assessment before communicating personal information outside Quebec. Several vendors offer Canadian data residency — firms should ask. Contingency and cloud guidance: most law societies' cloud-computing checklists (encryption, access control, breach terms, data return on termination) transfer directly to AI receptionist vendor selection.
Everything above reduces to a configuration and contract checklist:
Configured this way, an AI receptionist is not an ethics risk to manage down — it is frequently an ethics improvement: every call disclosed, every intake identical, every transcript reviewable, versus the unrecorded, unsupervised variability of a rushed human answering 60% of calls. More context across the topic lives on the AI Receptionists for Law Firms hub, and our services team configures all of the above as standard.
LexScale.ai designs, scripts, and deploys AI receptionists for law firms across Canada and the United States — configured for your practice areas, your practice management software, and your jurisdiction's ethics rules. See what your missed calls are costing you with our missed-call calculator, then book a call.
Book a Free Strategy Call →Free Tools & Resources
AI SEO for Law Firms · AI Website Design for Law Firms · Free AI Visibility Grader · AI for Every Practice Area
Ready to grow your firm with AI?