The Short Version
California's SB 574 takes effect January 1, 2027 and writes attorneys' use of generative AI into statute, so review your intake chatbot, AI receptionist and AI-written pages now. It bars entering confidential or personal identifying information into an AI system other people can reach. Clio found 82% of consumers want to know when their lawyer uses AI.
Your intake runs on software now. A chat window on the site answers at 11 pm, an AI voice agent picks up when the front desk is on another line, and a drafting tool writes the practice-area pages and the Google ads. California has now written the rules for that software into a statute, and the first date on the calendar is January 1, 2027.
Governor Newsom signed SB 574 on September 30, 2026. It is the first law in the country that tells attorneys how they may use generative AI, and it applies to law firms whose marketing team did nothing wrong under the old rules. SB 574 bars entering confidential or personal identifying information into an AI system that other people can reach, demands verification of every output, and tells you to consider disclosing AI use in anything you create for the public. On October 26, Clio's 2026 Legal Trends Report put a number on the other side of the desk: 82% of consumers want to know when their lawyer uses AI.
California law firms can fix their intake and advertising before January, and the ones that do will pick up the clients the others lose. Below: what SB 574 and the State Bar's 2026 guidance say in plain words, the six places they hit a California law firm's intake and advertising, how New York and Texas treat the same tools, and how to disclose AI use so that it raises client trust.
What California changed for law firms in the fall of 2026
California made three things law on September 30, 2026: an attorney may not delegate the practice of law to generative AI, may not put confidential or personal identifying information into an AI system without restricted access, and must verify AI output, disclose AI use to the court, and consider disclosing it in content created for the public. All of it takes effect January 1, 2027.
SB 574 in plain words
SB 574 adds section 6068.1 to the Business and Professions Code, and the chaptered text (Chapter 858) is short. An attorney who uses generative AI to assist in the practice of law must not enter confidential, personal identifying, or other nonpublic information into a system unless access to that information is restricted to the attorney and people the attorney has authorized under confidentiality obligations. The attorney must take reasonable steps to verify outputs, including every citation, and correct any erroneous or hallucinated output. The attorney must disclose AI use to the court for all documents submitted, and must consider whether to disclose it when AI is used to create content provided to the public.
The definition of personal identifying information is the part that reaches your website. It includes dates of birth, Social Security numbers, driver's license numbers, the addresses and phone numbers of parties, victims and witnesses, medical or psychiatric information, financial information and account numbers. An injury or family law intake form asks for most of that list. The bill also amends section 128.7 of the Code of Civil Procedure so that no paper filed in a California court may contain a citation the responsible attorney has not personally verified.
The State Bar's 2026 guidance, adopted May 14
The State Bar of California's Board of Trustees approved the updated Practical Guidance for the Use of Generative Artificial Intelligence on May 14, 2026, replacing the 2023 version, per the Board's agenda item.
The agentic section of the guidance matters most to a firm's marketing. The guidance lists "autonomously facilitating client intake" as an example of an agentic workflow. A lawyer must not deploy agentic systems in a manner that lets the system communicate legal advice or act in a representative capacity without meaningful lawyer supervision, and must not permit autonomous external transmission of client information, including automated communications, without safeguards and human review. Under Rule 1.4, a lawyer must consider disclosing to the client that they intend to use AI, how, and with what risks.
Proposed rule comments are a third track. The Committee on Professional Responsibility and Conduct proposed comment amendments to Rules 1.1, 1.4, 1.6, 3.3, 5.1 and 5.3 on March 13, 2026, including a comment that "reveal" in the confidentiality rule covers exposing confidential information to AI tools, and AI Law Librarians reports the committee voted on September 11, 2026 to send them to the Board of Trustees. They bind only if the Board and the Supreme Court adopt them.
What clients told Clio on October 26
Clients already assume you use AI and want to be told. Clio's 2026 Legal Trends Report, released October 26, 2026, found that nearly one in four U.S. adults has used AI to answer a legal question, and 46% say they would consult AI before contacting a lawyer. On trust, 82% of consumers want to know when their lawyer uses AI, and 70% say they trust a lawyer more when that use is disclosed. Clio sells AI products, so treat the figures as a vendor's survey.
Where the new rules hit a California law firm's pipeline
The new AI rules hit a California law firm's pipeline in six places, and the first five are sharpest in California. Each one below ends with what to do before January 1, 2027.
Pain point 1: the chat widget that collects a date of birth
A law firm's intake chat already asks for the exact information section 6068.1 names. A Los Angeles injury prospect types the crash date, the insurer and the hospital into a widget at midnight. If the widget sends that text to a model whose access is not restricted to your firm, the attorney who set it up is on the wrong side of the statute from January 1. Baker McKenzie reads it the same way: an enterprise AI subscription alone does not establish compliance. What to do: get the vendor's data terms in writing and move the fields on the statutory list behind a human handoff.
Pain point 2: the receptionist that never says it is a bot
An intake bot that passes as a person is a misleading communication, and in California it is also a consumer statute. Business and Professions Code section 17941 makes it unlawful to use a bot to communicate with a person in California online with the intent to mislead them about its artificial identity in order to incentivize a purchase or sale of goods or services. A clear and conspicuous disclosure that it is a bot removes liability. The Oregon State Bar's Formal Opinion 2026-208 notes that more than one state has said a chatbot that does not identify itself could run afoul of the rule against misleading communications. What to do: make the first line of every chat and AI voice call a plain statement that the caller is talking to an automated assistant.
Pain point 3: the practice page nobody reviewed
Content "provided to the public" now carries a statutory duty to consider disclosure, and no firm has a definition of where that starts. Scott Meyers, chairman of Akerman LLP, told Bloomberg Law on October 2, 2026 that "the question every firm will be working through before January is what counts as using AI". A drafted practice page, a platform-suggested ad headline and a chatbot answer are all content created for the public. What to do: keep a review log for every AI-assisted page and ad, with the attorney who approved it.
Restricted access is the test under section 6068.1. The statute does not care whether a vendor markets its tool as "legal AI". It asks whether the confidential and personal identifying information your attorney enters can be reached by anyone outside the attorney and the people the attorney has authorized under a duty of confidentiality. A consumer chatbot plan fails that test on its own terms of service, and an enterprise deployment passes it only when the contract and the configuration show it does.
Pain point 4: the agent that sends the engagement letter
An intake agent that takes the next step on its own can create a client you never agreed to represent. The Oregon opinion walks through it: one agent gathers information, another drafts an engagement letter, and a third sends it, which a prospect may reasonably read as the start of an attorney-client relationship. Oregon adds that autonomous actions by a chatbot should be subject to lawyer approval, with a responsible lawyer automatically notified when an interaction occurs. What to do: put a human approval gate in front of every outbound message the agent drafts, and enter every prospect the bot screens into your conflict system.
Pain point 5: the brief your marketing AI touched
SB 574's disclosure duty to the court covers all documents, so from January 1 a firm must disclose AI use in any filing its drafting assistant touched. The National Law Review reported that in Noland v. Land of the Free, L.P., a California Court of Appeal sanctioned counsel $10,000 after finding that nearly all of the quotations in his briefs had been fabricated by generative AI. What to do: separate the marketing AI workspace from the litigation workspace, and keep a per-matter log of which filings AI touched, a trail WilmerHale notes the statute requires in no set form.
Pain point 6: one AI policy running in three states
A firm with offices in Los Angeles, New York City and Dallas is governed by a statute, an advertising disclosure law and an ethics opinion that do not line up. California regulates the attorney's input and output. New York regulates the AI-generated person in your video ad. Texas has no AI statute for private businesses and leans on a 2025 ethics opinion and a filing rule for ads. What to do: build one state-by-state matrix, like the one in the next section.
How the rules land in California, New York and Texas
California is the only one of the three with a statute aimed at attorneys' AI use, New York is the only one with an advertising disclosure law for AI-generated people, and Texas regulates the same conduct through its disciplinary rules and ethics opinions.
California: the statute arrives on January 1, 2027
A California firm has eight weeks to make its intake and content fit section 6068.1. For a firm in Los Angeles, San Francisco or San Diego, the practical order is this. List every tool that touches a prospect's information and get each vendor's data terms in writing. Mark which intake fields match the statutory list and route those to a person or a restricted-access system. Start the per-document AI log for anything headed to court, because the Noland sanction came under existing rules, before the statute added a personal verification duty.
Two California rules apply with no January wait. Section 17941, the bot disclosure law, already covers a chat widget on a Los Angeles firm's site. The State Bar's May 14, 2026 guidance is also the reference a managing partner is held to in a complaint today. The statute tells you to consider disclosure, and Clio says disclosure raises trust, so there is a local upside too. A San Diego immigration practice or a San Francisco employment firm that is first to put a plain AI statement on its intake page gets a trust signal out of a compliance line, a real edge in California legal marketing.
| State | Rule on the attorney's AI use | What your intake bot must do | What your ads must disclose | Date that bites |
|---|---|---|---|---|
| California | SB 574, Bus. & Prof. Code 6068.1: no delegation, restricted access for confidential and personal identifying information, verify outputs, disclose to court, consider public disclosure | Identify itself as a bot (Bus. & Prof. Code 17941); keep statutory personal identifying information out of unrestricted systems; no autonomous outbound messages without human review | Consider disclosure for any AI-created public content; keep a review log per page and ad | January 1, 2027 (17941 already in force) |
| New York | No attorney AI statute; Rules 7.1 and 7.3 plus New York City Bar Formal Opinion 2024-5 | Identify itself; keep the "Attorney Advertising" label on the home page | Conspicuous disclosure of any synthetic performer in a visual ad (General Business Law 396-b); disclosure of actors and fictionalized events (Rule 7.1); website archived every 90 days | June 9, 2026 (synthetic performer law in force) |
| Texas | Professional Ethics Committee Opinion 705 (February 2025): verify output, no confidential input without consultation and consent unless satisfied the tool protects it | Comply with the misleading-communication rules; no statute requires a private firm's bot to disclose itself | File ads with the Advertising Review Committee within 10 days unless exempt; website pages exempt except the home page | In force now |
New York and Texas: same tools, different tripwires
New York regulates the AI-generated person who appears in a law firm's ad, and leaves the lawyer's prompts to the ethics rules. The synthetic performer law, S8420A, amended General Business Law section 396-b and took effect June 9, 2026, per the bill text. An advertiser must conspicuously disclose that a synthetic performer, an AI-made likeness of a human who is not an identifiable real person, appears in the ad. The civil penalty is $1,000 for a first violation and $5,000 for each subsequent one, and audio-only ads are exempt. Kelley Drye notes the law prescribes no wording. For a New York City or Buffalo firm running AI-generated presenters in video spots, the disclosure goes on screen. Separately, under Rule 7.1 an ad may not use actors to portray the lawyer or clients, or depict fictionalized events, without disclosure. The New York City Bar's Formal Opinion 2024-5 is the authority Oregon cites on unidentified chatbots, so a New York firm's bot should introduce itself too.
Texas has no AI statute for law firms, so the duties sit in Opinion 705 and the advertising filing rules. The Professional Ethics Committee for the State Bar of Texas issued Opinion 705 in February 2025. A lawyer not reasonably satisfied that a generative AI program keeps confidential information from others should at a minimum not input any without client consultation and consent, and should always verify the output. A Dallas or Houston firm's intake bot is measured against that standard. Texas's own AI law, in force since January 1, 2026, puts its consumer disclosure duty on governmental entities, per DLA Piper's summary, so a private firm's bot disclosure is a disciplinary-rule question. Under Rule 7.04, a lawyer must file an advertisement with the Advertising Review Committee no later than 10 days after dissemination. Rule 7.05 exempts website content except the home page. An AI-written home page for an Austin firm is a filing; an AI-written blog post on the same site is not.
How to disclose AI use without losing the client
A law firm should disclose AI use where the client is deciding, say what the human does, and write the statement in the firm's own voice. The Clio data supports this: 82% of consumers want to know when their lawyer uses AI, 70% trust a lawyer more when it is disclosed, and only 6% are comfortable with AI in high-stakes personal matters.
Say it where the client is deciding
The AI disclosure belongs on the intake page, in the first line of the chat, and in the engagement letter. A privacy policy footer is too late in the client's decision. A line such as "Our online assistant is automated. A licensed attorney reviews every inquiry before we contact you, and we do not place your medical or financial details into public AI tools" answers the section 6068.1 question and the Clio question in the same breath.
Keep the human visible in every AI-assisted touch
Name the attorney who reviewed each page or ad. The California guidance keeps the lawyer fully responsible for AI output, and Rule 1.4 asks the lawyer to consider telling the client how AI is used. Clients want to know who is accountable, and the statute and the Clio survey both point to a named lawyer as the answer.
A well-placed AI disclosure can help conversion. Clio's October 26, 2026 report says 70% of consumers trust a lawyer more when AI use is disclosed. A firm that writes "automated assistant, attorney-reviewed" at the top of its intake flow is answering the question 82% of prospects are already asking.
What works: intake and content that pass the statute and the survey
Three things pass both SB 574 and the Clio survey: restricted-access intake proven by contract and configuration, a bot that introduces itself and hands off to a named person, and a review ledger that records who checked every AI-assisted page, ad and filing.
Restricted-access intake, by contract and by configuration
Restricted access is proven in two documents: the vendor's data terms and your own configuration record. The terms must say that your inputs are not used to train models, are isolated from other customers, and are deleted on a schedule you control. The configuration record must show which intake fields the AI touches and which go straight to a person. Date the review and repeat it when the vendor changes its product.
A bot that introduces itself and hands off
A law firm's intake bot should identify itself, collect the non-sensitive facts, and get a human on the line. Its first line says it is automated. It then hands off to a person inside a stated window and notifies the responsible lawyer automatically. Keep the statutory list out of the bot until a person is on the line or the data lands in a restricted-access system.
An AI content ledger for ads, pages and filings
An AI content ledger needs three columns: what the AI produced, who reviewed it, and where it was published or filed. For public content the ledger answers the "consider whether to disclose" duty with a record of the decision. For filings it is the per-document trail the statute requires but does not define.
How Leadnox helps California law firms get ahead of January 1
Leadnox rebuilds a law firm's intake, website content and ad program so every AI-assisted touch has a disclosure, a human reviewer and a data path that meets section 6068.1, then measures whether the change lifts consultation bookings.
Intake rebuilt around section 6068.1
We map every tool that touches a prospect against the statutory list of personal identifying information and the restricted-access test. Sensitive fields move behind a human handoff or into a configuration that passes the test, and the bot gets its introduction line and approval gate. The result is an intake flow a managing partner can defend and a prospect can trust at midnight.
Content and ads with a disclosure that reads as confidence
For content and ads, we write the AI disclosure in the firm's voice, place it where prospects decide, and attach a named reviewer to every practice page and ad. For firms advertising in New York we add the synthetic performer disclosure and the Rule 7.1 dramatization notice to the creative brief. For Texas we file with the Advertising Review Committee inside the 10-day window.
One matrix for California, New York and Texas
Multi-office firms get one state-by-state matrix, owned by one person per office, covering the attorney's AI use, the bot's behaviour and the ad disclosures. See how we approach marketing for law firms if you want the intake audit before the statute takes effect.
Measuring whether the change paid for itself
A California law firm can measure the change with four numbers: the consultation rate after the bot introduces itself, the share of leads handled under restricted access, ledger coverage of AI-assisted content and filings, and the trust signal in client feedback. A good result is a flat or rising consultation rate, full coverage on the second and third measures, and prospects who name a person.
Consultation rate after the disclosure line
Compare the share of chats and calls that become booked consultations in the four weeks before and after the "automated assistant, attorney-reviewed" line goes live. When the rate drops, the usual cause is a handoff window that is too long. The disclosure line itself is rarely what turns a prospect away.
Share of leads under restricted access
Count the inquiries whose sensitive fields were collected by a person or by a system that meets the restricted-access test, divided by all inquiries. The target is every inquiry from January 1, 2027.
Ledger coverage
Count AI-assisted pages, ads and filings with a reviewer and a disclosure decision recorded, divided by all AI-assisted items. The target is every item.
Trust signals in reviews and intake notes
Track how often prospects mention the automated assistant, the named attorney, or the speed of the human callback in reviews and intake notes. Comments that name a person are an early sign that the disclosure is working.
Frequently Asked Questions
Yes, where an attorney uses it to assist in the practice of law. From January 1, 2027, section 6068.1 bars entering confidential or personal identifying information, such as dates of birth, phone numbers of parties, and medical or financial details, into a generative AI system unless access is restricted to the attorney and authorized people. A chatbot that routes those fields into an unrestricted model fails that test, so move sensitive fields behind a human handoff.
In practice, yes. Business and Professions Code section 17941 makes it unlawful to use a bot to mislead a person in California online about its artificial identity to incentivize a purchase of services, and a clear, conspicuous disclosure that it is a bot removes liability. Ethics opinions in other states also treat an unidentified chatbot as a potentially misleading communication. Put the disclosure in the first line of every chat.
A conspicuous disclosure in the ad that a synthetic performer appears in it. New York's General Business Law section 396-b, amended by S8420A and in force since June 9, 2026, covers any AI-generated human likeness that is not an identifiable real person, with a civil penalty of $1,000 for a first violation and $5,000 for each later one. Audio-only ads are exempt. Rule 7.1 separately requires disclosure of actors and fictionalized events.
Not unless the lawyer is reasonably satisfied the tool will not reveal confidential information to others. Texas Professional Ethics Committee Opinion 705, issued in February 2025, says a lawyer who is not satisfied should at a minimum not input confidential information without client consultation and consent, and should always verify the tool's output. What a prospect types into an intake bot is confidential information, so the same standard applies to the vendor behind the bot.
The available data says it is more likely to help. Clio's 2026 Legal Trends Report, released October 26, 2026, found 82% of consumers want to know when their lawyer uses AI and 70% trust a lawyer more when that use is disclosed, while only 6% are comfortable with AI in high-stakes personal matters. A disclosure that names the human reviewer answers both the trust question and California's new duty to consider disclosing AI-created public content.
Sources
- The chaptered text. leginfo.legislature.ca.gov
- The Board's agenda item. calbar.ca.gov
- The guidance. calbar.ca.gov
- AI Law Librarians reports. ailawlibrarians.com
- Clio's 2026 Legal Trends Report. clio.com
- Reads it. bakermckenzie.com
- Section 17941. california.public.law
- Formal Opinion 2026-208. osbar.org
- "the question every firm will be working through before January is what counts as using AI". news.bloomberglaw.com
- Reported. natlawreview.com
- Notes. wilmerhale.com
- The bill text. nysenate.gov
- Notes. kelleydrye.com
- Rule 7.1. law.cornell.edu
- Opinion 705. legalethicstexas.com
- Summary. dlapiper.com
- Rule 7.04. legalethicstexas.com
- Rule 7.05. legalethicstexas.com
Drafted with AI assistance from the public sources listed above. How we write. This article is general information for business owners, not legal, tax, medical or financial advice. Rules change and differ by state, so check the source documents or ask a qualified professional before you act on it.
Is your California intake ready for section 6068.1?
Leadnox rebuilds intake, website content and ad programs for law firms in California, New York and Texas so every AI-assisted touch has a disclosure, a human reviewer and a compliant data path. Start with a free intake audit.