Safe AI use by lawyers no longer rests on cautious prompts but on a demonstrable control layer per matter: assess each tool beforehand on data practices, always treat AI output as a draft under your own responsibility, inform clients and record per case which tool processed which data and who approved the output.
That shift is not optional. The American Bar Association issued Formal Opinion 512 in 2024, the first national ethics opinion on generative AI in the legal profession, and in July and August 2026 concrete pieces followed from among others the Alabama State Bar (reported by Reuters) and the New York City Bar Association. Together they show that AI use falls under existing professional duties and that a firm must be able to demonstrate how it complies with them. In our assessment, a comparable logic applies to notaries, because they too deal with professional secrecy, due care and identification.
What does ABA Formal Opinion 512 change about lawyers' duties regarding AI?
According to Epiq's analysis, ABA Formal Opinion 512 explicitly places generative AI under the existing rules of conduct. A lawyer need not be an AI expert, but must have a reasonable understanding of the capabilities and limitations of the specific tool they use. The opinion links AI directly to familiar core duties:
- competence: understand what the tool can and cannot do;
- confidentiality: assess data practices before client information is entered;
- communication: inform clients about AI use when they ask or when arrangements require it;
- candour towards the court: check output for accuracy before it enters a matter or court document;
- supervision and reasonable fees: record firm policy and oversight.
The practical consequence: AI use is no longer a grey area but a regular part of professional ethics. Anyone wishing to comply with these duties must set up AI in professional practice responsibly rather than using a public chatbot ad hoc.
Why may you not bill AI-saved hours according to the Alabama State Bar?
Reuters reported on 23 July 2026 that the Alabama State Bar updated its ethics guidance. The core: lawyers may not bill hours saved thanks to AI, but they may bill the time they spend reviewing, correcting and applying AI-generated work with professional judgement.
That places the value exactly where the professional rules want it: with human verification. AI output remains under the lawyer's responsibility, and fact-checking is not an extra service but a condition for complying with the rules of conduct. As an editorial observation: this calls for time recording and firm policy in which the verification step is visible and accountable, not hidden behind a productivity claim.
Which questions must you answer per AI tool before entering client data?
Arjun Jaggi's analysis of deploying language models under professional secrecy translates ABA 512 into concrete questions you must be able to answer before use. Treat this as a fixed assessment list per tool:
- Does the tool use entered client data to train models?
- How long are prompts and documents stored?
- Who has access to that data, and under which jurisdiction?
- Is that compatible with professional secrecy and the applicable rules of conduct?
- Does the contract with the supplier record these points?
This assessment should be recorded verifiably, not agreed orally. For the trade-off between local and cloud processing, it helps to determine per task what you need to audit; we addressed this earlier when having AI process confidential documents without unnecessarily disclosing the content.
How do transcription, summarisation and notarial drafts fit these requirements?
The New York City Bar Association addresses in Formal Opinion 2026-2 the use of AI for recording, transcribing and summarising conversations with non-clients. The opinion emphasises that lawyers must protect confidentiality and privilege, must safeguard consent and transparency around AI recording, and that the choice of a service depends partly on the provider's security and data processing.
This shows that even seemingly neutral applications touch on confidentiality. In our assessment, the same applies to notarial practice: AI drafts for deeds, summaries or due diligence can be useful, provided the tool used has been assessed beforehand and human control remains central.
How do you translate these bar opinions into a control layer per matter?
Legal AI Compliance's 50-state tracker shows that more and more bars address the same themes: technological competence, protection of client confidentiality, verification of output and clear firm policy with supervision. ABA 512 serves as a national reference point in this. Our editorial translation into a workable architecture:
- a firm-wide AI policy recording which tools are permitted for which tasks;
- a mandatory assessment of data practices and contracts per tool, recorded verifiably;
- a workflow rule that AI output is always a draft and is verified by a lawyer or notary before it enters the matter;
- logging: who used which tool on which matter information, and who approved the output.
Anyone who designs this layer can show to a client, professional body or supervisor how AI was deployed in a case. For keeping AI controllable across separate tools, we described earlier how to keep AI workflows in legal matters controllable, and why a separate a verification layer for legal AI research tools remains necessary.
In this design, a verification console such as IamVera.ai can be positioned as a visibility and control layer above the tools. Vera is not a chatbot and not its own language model, but can route a task through selected independent models and make verification steps, corrections and sources visible for inspection. The Semantic Privacy Shield can replace sensitive values before processing on EU infrastructure with synthetic, session-only equivalents; if the privacy check fails, the document is not sent onward. This supports making safe usage patterns demonstrable, but the professional final judgement always remains with the lawyer.
Sources and references
- American Bar Association Ethics Ruling on Generative AI
- Alabama updates ethics guidance for lawyers amid worries about AI misuse
- Formal Opinion 2026-2: Ethical Use of AI for Recording, Transcribing and Summarizing Non-Client Conversations
- State Bar AI Ethics Opinions: The Free 50-State Tracker (2026)
- Deploying LLMs Under Attorney-Client Privilege and Bar Ethics Rules
Sources: The article draws on ABA Formal Opinion 512 (via Epiq), Reuters on the Alabama State Bar, NYC Bar Formal Opinion 2026-2, the 50-state tracker from Legal AI Compliance and a privacy analysis by Arjun Jaggi.