Why Smart Lawyers Forward Instead of Uploading

The dangerous moment is not the first time you use generative AI. The dangerous moment is the first time you paste a client letter into a consumer chat window because you are late and the interface feels “private enough.” Competence is not the issue. Convenience is. The bar’s question is not whether models are powerful. It is whether your workflow preserves duties while you use them.

Where do lawyers accidentally breach duties with AI?

Lawyers most often breach duties when confidentiality, competence, and candor get lost in the gap between “helpful output” and “controlled workflow.” The ABA’s Formal Opinion 512 on generative tools—see the <a href="https://www.americanbar.org/content/dam/aba/administrative/professional_resolution/formal_opinion_512.pdf" target="_blank" rel="noopener noreferrer">ABA Formal Opinion 512 PDF</a>—frames competence, confidentiality, and communication obligations in language any partner can cite in a staff meeting. California’s practical guidance on generative AI—<a href="https://www.calbar.ca.gov/Attorneys/Conduct-Discipline/Rules/Rules-of-Professional-Conduct/Generative-AI" target="_blank" rel="noopener noreferrer">State Bar of California generative AI resources</a>—reminds attorneys that confidentiality travels with every paste into a consumer product. via.email reduces the paste temptation by keeping help mail-shaped: you forward what you choose to Explain Legal Letter at explain.legal.letter@via.email or Summarize Contract Obligations at summarize.contract.obligations@via.email, and you keep the reply in a thread you can search later—without granting inbox access or send-on-your-behalf automation.

Courts and commentators are not theoretical here. Reuters documented how <a href="https://www.reuters.com/legal/michael-cohen-said-he-unwittingly-sent-his-attorney-fake-legal-cases-generated-by-ai-2024-05-31/" target="_blank" rel="noopener noreferrer">fake cases generated by AI can reach real filings</a> when verification fails. The Federal Judiciary’s guidance on generative AI for judges—<a href="https://www.uscourts.gov/about-federal-courts/court-website-policies/judiciary-policy-on-generative-ai-in-the-courts" target="_blank" rel="noopener noreferrer">US Courts policy page</a>—signals that docket integrity is under active scrutiny. MIT Technology Review’s 2026 reporting on whether a <a href="https://www.technologyreview.com/2026/02/11/1132768/is-a-secure-ai-assistant-possible/" target="_blank" rel="noopener noreferrer">secure AI assistant is possible</a> is the technical mirror: assistants that touch sensitive content inherit sensitive-content risk.

What does ABA Opinion 512 emphasize for everyday practice?

Opinion 512 emphasizes what good lawyers already know in their bones: you remain responsible for output you file or send, you must protect client information, and you must communicate honestly about how work was prepared when it matters. The opinion is not anti-tool. It is pro-competence. The failure mode is speed without custody: a draft that sounds authoritative, a citation that never existed, a workflow nobody can reconstruct when a grievance arrives. via.email supports conservative workflows by keeping tasks bounded—Redline Contract Version at redline.contract.version@via.email works on the materials you forward, and you keep human review where the rules require it.

Why are consumer chat upload paths higher risk than deliberate forwarding?

Upload paths encourage maximal context dumping. Email forwards, done with discipline, can be minimal: the clause, the question, the redaction strategy. The difference is not moral superiority. It is operational control. When you upload an entire PDF because the UI nudges you toward “more context,” you increase the surface area of accidental disclosure. California’s guidance is explicit that confidentiality duties travel with those choices. via.email does not access your mailbox or cloud drives; you initiate each task, which is closer to the “conscious delegation” pattern firms prefer on paper.

Gartner’s March 2025 commentary—<a href="https://www.gartner.com/en/newsroom/press-releases/2025-03-25-gartner-says-cfos-should-reset-expectations-about-ais-impact-on-workforce-productivity-and-headcount" target="_blank" rel="noopener noreferrer">CFOs resetting AI productivity expectations</a>—also matters for law firm economics: pressure to adopt quickly without a team-level ROI story is how shortcuts become policy accidents.

How do specialized legal agents change review economics?

Specialized agents change economics by shrinking the mechanical hours around review—summaries, first-pass obligation maps, structured comparisons—without pretending the lawyer is optional. Summarize Contract Obligations at summarize.contract.obligations@via.email can extract a checklist from a forward you control. Redline Contract Version at redline.contract.version@via.email supports comparison work when you supply versions. Explain Legal Letter at explain.legal.letter@via.email helps translate dense correspondence into plain language for client updates—still with your judgment on what may be shared. via.email is an email-based AI agents platform with Legal department agents; browse more at https://www.via.email/agents.

The economic point is not “bill fewer hours.” It is “spend hours on judgment, not on retyping the same structural task for the hundredth time.”

What is a conservative workflow pattern partners can approve?

A conservative pattern has four steps. First, classify the material: client secrets, work product, public filings—different rules apply. Second, choose the smallest forward that still asks a real question. Third, use a specialist address tied to a narrow job, not a general chat. Fourth, verify anything that could reach a court or a client: citations, dates, quotes, and “facts” implied by confident prose. McKinsey’s <a href="https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-social-economy" target="_blank" rel="noopener noreferrer">interaction worker research</a> is the blunt reminder that communication labor still consumes huge weekly hours for knowledge workers—including lawyers—so workflow design matters as much as model choice.

What should you tell clients when you change process?

Tell clients the truth in plain language: you may use AI-assisted drafting or analysis on tasks you define, you remain responsible for final work product, and you will not input confidential information into tools your firm has not vetted. The goal is informed consent without scaring people with sci-fi. If you need internal language first, email Frame AI Adoption at frame.ai.adoption@via.email with stakeholder concerns; you get calmer memo drafts you can adapt to firm policy—still requiring human sign-off.

How should associates log prompts and outputs for conflicts checks?

Logging should be boring and searchable: matter number, date, what was sent (at a high level), what was received (stored as a reply), and who approved use in the file. The point is reconstructability. A consumer chat tab that disappears is the opposite of a conflicts-ready record. via.email helps because outputs return as email replies you can file like any other work product—still subject to your firm’s retention rules.

Harvard Law School’s Center on the Legal Profession—<a href="https://clp.law.harvard.edu/" target="_blank" rel="noopener noreferrer">CLP at Harvard</a>—is a useful neutral link for firm-management context without vendor spin.

Law.com and Bloomberg Law remain standard practitioner reading environments for ongoing AI policy coverage—use them the way you use any other secondary source: verify, do not outsource thinking.

The honest close

The best lawyers were never going to “win” by avoiding AI. They were going to win by keeping custody—of client confidences, of court credibility, of their own attention. Forwarding instead of uploading is not a magic shield. It is a discipline: smaller surface area, initiated by you, auditable later. via.email is one mail-native way to run that pattern with Legal specialists at explicit addresses.

Related reading: Lawyers Spend 2.5 Hours a Day on Email. Precision AI Without the Dashboard., Disclosure Still Breaks on Paste. Email Is Where Fixes Live, Contract Deadlines Hide in Attachments. Surface Them Faster, and Lawyers Spend 66% of Day on Email, 2.5 Hours Billable.

If you want a partner-level stress test for any new AI habit, ask one question: “If this output were wrong, how would we prove what we did?” If the answer is “we cannot,” you do not have a workflow—you have a gamble dressed up as innovation. The forward-and-reply pattern is not perfect, but it is closer to evidence than a browser session nobody archived.

Power is not the hard part anymore. Custody is.