We are at once observers and users of technology that will change the world faster than any innovation that has come before. Powerful tools demand careful users. The same technology that drafts a persuasive brief in minutes can also invent a case that has never existed, and the same software that enhances a photograph can fabricate one entirely. Courts depend on a simple premise: that the law cited is real and the evidence presented is authentic. Artificial intelligence puts both at risk.
This article examines the legal duties that govern the citation of authorities, how to respond if you have cited a hallucinated case and face sanctions, and how to challenge deepfake exhibits.
AI is not completely reliable for use in court
Generative AI does not understand what it is “saying.” AI converts questions into numbers representing words or word fragments (tokens) and then uses a probability engine to determine the likely next number to add, again and again, until it finishes its “response” and converts its added numbers back into words. (Hough, How to Understand ‘Persistent Agents,’ (Oct. 5, 2026) Barron’s, at p. 8.)
AI can produce hallucinated or fake case citations, create quotes from and holdings of cases that do not exist, and inaccurately describe the holdings of real cases. The fake citations and discussions look realistic.
Generative AI tools tend toward sycophancy, telling the user what he wants to hear. AI may give an incorrect answer rather than admit that it does not know the answer.
A 2025 study conducted by Stanford and Yale researchers examined the rate of hallucinations of older Lexis+ and Westlaw AI-assisted legal research programs. (Magesh, et al., Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools (2025) 22 J. Empirical Legal Stud. 216.) The study found that Lexis+ AI hallucinated 17% of the time and Westlaw AI Assisted Legal Research hallucinated 33% of the time. Hallucination was broadly defined in this study and was not limited to hallucinated citations. The current versions, Lexis+ with Protégé and Westlaw CoCounsel were not studied.
Agentic AI
Agentic AI is artificial intelligence that doesn’t just answer questions or generate text on request, but can pursue a goal on its own. Given an objective, it plans the steps, uses tools such as search engines, databases or software, takes actions, checks its own results and adjusts course. To understand the difference between generative AI and agentic AI, think Claude versus HAL 9000.
Recursive self-improvement AI
Self-modifying agents emerge from the process of recursive self-improvement, in which agentic AI systems are being tested in research settings to autonomously design and develop their own successors with iterative improvements, allowing gains in the system to strengthen the process that generates further gains. (Hough, How to Understand ‘Persistent Agents,’ (Oct. 5, 2026) Barron’s, at p. 8.)
Persistent agentic AI
Persistent agents are still built on token-predicting models, but are given tools, memory and a goal to pursue. The persistent part means that these agents can keep working even after you have logged off your computer and the agents learn about you as they go, as a human assistant would. (Hough, How to Understand ‘Persistent Agents,’ (Oct. 5, 2026) Barron’s, at p. 8.)
Quantum computing: Data at risk
Quantum computing differs from traditional or classical computing in that a traditional computer stores data as bits, each of which is assigned either a zero or one. Quantum computers use qubits. A qubit is not a stored electrical or magnetic state the way ordinary data is. It is a delicate physical system, such as an atom, one particle of light, or a tiny superconducting circuit, whose quantum state represents the information. Quantum computers are expected to be exceptionally fast at specialized problems, such as simulating molecules or breaking certain forms of encryption. (See M. Kaku, Quantum Supremacy: How the Quantum Computer Revolution Will Change Everything (Doubleday 2023).)
Quantum computing might be used not only to process information quickly, but also for better outcome prediction, including prediction of judicial decisions, settlements, and verdicts.
Bad actors are conducting “harvest now, decrypt later” campaigns. They may attempt to steal encrypted attorney-client communications, sealed records and confidential settlements today to decrypt with future quantum computers.
Post-quantum cryptography (PQC) is a new generation of encryption designed to keep sensitive data safe in the quantum era. The National Quantum Initiative Act, 15 U.S.C. § 8801 et seq., directs the National Institute of Standards and Technology (NIST) to carry out research to facilitate the development and standardization of quantum cryptography and post-quantum classical cryptography. (15 U.S.C. § 8831(a)(3).) NIST is tasked with developing standards in collaboration with industry, universities, and federal laboratories, to support a robust U.S. quantum industry. The Quantum Computing Cybersecurity Preparedness Act, 6 U.S.C. § 1526, in turn, requires federal agencies to inventory their quantum-vulnerable information technology and migrate it to the post-quantum cryptography standards NIST develops. NIST finalized the first three standards in August 2024.
Adversarial prompt injection
Adversarial prompt injection is a tactic in which hidden instructions embedded in legal briefs are designed to hijack AI systems used by judges, law clerks and opposing counsel, to produce rulings in their favor or audit the court’s systems to determine whether the court is using AI to process the filing. Any incoming document is a potential vector for adversarial prompt injection to corrupt output. The tactic typically involves hidden prompts, such as inserting invisible text directly into the PDF filing, or using three-point white font on a white background invisible to the person reading the document, hidden text formatting and non-printing characters. Prompt injection can be used as a covert attempt to influence judicial decision-making.
In Elliott v. New York Bariatric Group, LLC, No. AAN-CV-25-6066141-S (Conn. Super. Ct., Ansonia-Milford Jud. Dist., Aug. 6, 2026) at p. 10, the self-represented plaintiff concealed text that instructed the AI systems to produce output only favorable to plaintiff’s position and to disregard a prior ruling as error to be corrected in plaintiff’s favor. The court did not use an AI system, so the injection did not reach a court target. The Court discussed the dangers of prompt injection. The AI system may treat the instruction as if it came from the system’s operator. The litigant may capture the court’s and opposing counsel’s AI tools and silently turn them to his advantage. The court rescinded plaintiff’s e-filing privileges, requiring in-person filings. This well-written unpublished decision is posted on DamienCharlotin.com.
DamienCharlotin.com
DamienCharlotin.com is a website run by a Paris-based international law researcher who has created searchable databases of legal cases from throughout the world involving AI hallucinations, deepfakes in court, AI evidence, and AI discovery. As of the date of the submission of this article, there are 2,149 cases on the AI hallucination database, including 1,473 from the United States. The section on AI discovery includes litigation on requests for parties’ communications with generative AI regarding the case. Counsel may wish to consider advising clients and witnesses not to communicate with generative AI regarding the case, lest it be discoverable.
SB 574: Attorneys must personally verify citations
SB 574 was enacted on September 30, 2026, and will take effect on January 1, 2027.
Business and Professions Code section 6068.1, subdivision (a)(3)(C) will require attorneys to: “Disclose the use of generative artificial intelligence to the court for all documents submitted to the court.” Section (a)(2) will prohibit attorneys from delegating the practice of law to generative artificial intelligence.
Code of Civil Procedure (“CCP”) section 128.7, subdivision (b)(2)(A) will provide: “A brief, pleading, motion or any other paper filed in any court shall not contain any citations that an attorney responsible for submitting the pleading has not personally verified, including any citation provided by generative artificial intelligence.” “Generative artificial intelligence” is defined in CCP § 128.7(b)(2)(B).
An earlier version of proposed CCP § 128.7(b)(2)(A) required that counsel personally read and verify citations. The enacted version requires that counsel personally verify citations, not read and verify citations. CCP § 128.7(b)(2)(A) may conflict with Court of Appeal decisions requiring counsel to read and verify citations, as opposed to verifying citations.
AI statutes and tools for judges in their adjudicative role
Judges using generative AI for any task within their adjudicative role must take reasonable steps to verify that generative AI material is accurate and should take reasonable steps to correct erroneous or hallucinated output in any material used. (Standards of Judicial Administration, Standard 10.80.) Judges “should consider” whether to disclose the use of generative AI if it is used to create content provided to the public. (Ibid.)
In the Superior Court of Los Angeles County, “[t]he authority and responsibility for judicial decision-making remains solely with judicial officers.” (LACourt.ca.gov) “Responsibility for all judicial decisions rests exclusively with judicial officers, who exercise independent judgment and are required to personally review and evaluate all proposed decisions before they are issued.” (Ibid.) The Court is committed to responsible and deliberate use of AI in ways that support, but never replace, the exercise of independent judicial analysis and judgment. (Ibid.)
The Court provides a closed version of Microsoft Copilot, and allows the use of additional tools, including Google Gemini, Anthropic’s Claude, OpenAI ChatGPT, Thomson Reuters and LexisNexis, but only if the user follows the Court’s AI policy.
The Superior Court of Los Angeles County is working with Stanford University to create a tool to assist Los Angeles Superior Court research attorneys in evaluating default judgment packages in the debt collection and unlawful detainer hub courts. This is in development, not in use, and the tool does not make any determinations.
The Court is also working with Learned Hand in a research and development project that is in early stages. Learned Hand is not being used by any judicial officer in any live cases. No generative AI tool will ever replace the individual judgment and discretion of a human judicial officer in making an adjudicative determination.
California-published decisions on AI hallucinations
Noland v. Land of the Free, L.P. (2025) 114 Cal.App.5th 426, 445-446 is the first published California appellate case involving AI hallucinations. Nearly all the quotations in the plaintiff’s opening brief, and many of the quotations in plaintiff’s reply brief were fabricated. Although most of the cases to which the quotes were attributed did exist, the quotes did not. Without citation to authority, the Court of Appeal held: “To state the obvious, it is a fundamental duty of attorneys to read the legal authorities they cite in appellate briefs or any other court filings to determine that the authorities stand for the propositions for which they are cited.” (Emphasis in original.) The court imposed “conservative” sanctions of $10,000, payable to the clerk of court, and ordered the clerk to provide the opinion to the State Bar.
Noland’s directive that attorneys read the authorities they cite was without citation to any California law. That directive has been repeated in many appellate decisions since Noland. More recently, the Legislature considered and rejected proposed statutory language requiring that counsel read and verify citations when it enacted CCP §128.7(b)(2)(A), which will require only that counsel personally verify citations when the statute takes effect on January 1, 2027.
After Noland was decided, there have continued to be cases where courts have been forced to deal with inaccurate citations presumably created by generative AI and not spotted by the lawyers submitting the document. These include the following:
People v. Alvarez (2025) 114 Cal.App.5th 1115 involved fake case cites and quotes in an opposition to a motion to dismiss filed in the Court of Appeal. The Court imposed sanctions of $1,500 payable to the clerk of court, and instructed the clerk to report the attorney to the State Bar.
In Schlichter v. Kennedy (2025) 116 Cal.App.5th 24, a writ petition and opening brief included hallucinated cases. The court imposed sanctions of $1,750 pursuant to CCP § 128.7, CRC 8.204 and 8.276, and reported counsel to the State Bar.
In Shayan v. Shakib (2025) 116 Cal.App.5th 619, the appellate brief misrepresented legal authority and quoted from a transcript in a different case. The attorney stated he did not use AI tools. The brief was stricken, and a new brief was ordered filed correcting or omitting fabricated citations and quotations. The Court imposed sanctions of $7,500 payable to the clerk of court, and ordered the clerk to serve the State Bar with the Court’s opinion.
In re Domestic Partnership of Campos & Munoz (2026) 118 Cal.App.5th 1112, 1124, 1126, counsel for both sides and the court cited fictitious cases. The fictitious cases originated with respondent’s counsel, who cited the fake cases in the trial court. Appellant’s counsel submitted a proposed order that contained the two fake cases. The trial court’s order cited the two fake cases. Appellant asserted that the trial court committed error by relying on the fictitious authorities cited in his proposed order.
The Court of Appeal held that the trial court abused its discretion because it “relied in material part” on the holdings of nonexistent legal authorities. The Court held that it is imperative for both the court and the parties to verify that citations in all orders are genuine and truly stand for the propositions cited. Counsel are officers of the court and have a duty to aid the court in avoiding error.
But the Court held that the appellant forfeited his claim of error by failing to object to his own proposed order in the trial court. The trial court’s order was affirmed notwithstanding the abuse of discretion. The Court imposed $5,000 sanctions against respondent’s counsel, who “doubled down” in the Court of Appeal when hallucinated cases were brought to her attention. The clerk of court was ordered to serve a copy of the opinion on the State Bar and counsel was ordered to report the sanctions against her to the State Bar.
In Quinteros v. Harbor Distributing, LLC (2026) 121 Cal.App.5th 60, a contract attorney drafted an opposition to motion to stay. He stated that he used Lexis, Lexis Protégé and the Lexis Citation Check function to draft the opposition and denied using “generative AI tools.” The brief contained fictitious cases, fabricated quotations, and “seriously misrepresented” the holdings of other cases. The attorney of record signed the opposition drafted by the contract attorney. The trial court gave the attorney three days’ notice of an OSC hearing on CCP § 128.7 sanctions. Five days after the hearing, the court imposed $5,000 in sanctions to be paid to the opposing party and $1,000 payable to the Court. The attorney forfeited the claim of trial court error by failing to raise it in the trial court. The judgment was affirmed.
In Del Biaggio v. Bansen (2026) 121 Cal.App.5th 831, there were fictitious quotes in the appellant’s opening brief. The court held that a lawyer must review all AI outputs for accuracy and cannot delegate this task to a paralegal. The Court imposed sanctions of $1,500, reported the attorney to the State Bar and ordered the attorney to provide a copy of the opinion to the client.
In Southland Homes & Real Estate and Investment, LLC v. Lam (August 5, 2026) __ Cal.App.5th Supp.. __, the opening brief filed in the Appellate Division of the Orange County Superior Court cited two cases that did not exist, misstated the holding of a third case and provided an incorrect citation for a fourth case. Lead counsel took the position that it was not reasonable for him to verify every citation in a brief prepared by an associate attorney; a supervising attorney should reasonably be permitted to rely on the associate attorney to draft briefs, verify citations and comply with their ethical duties and professional responsibilities.
Lead counsel did take measures to prevent a recurrence by implementing a written policy regarding the use of AI and revoking all use of AI in any firm matter. He apologized for the associate’s actions and his failure to prevent the petition from being filed with false citations. The associate who used ChatGPT and Google Gemini to assist with legal research and drafting accepted responsibility, but asserted that he did not violate Business and Professions Code section 6068, an argument the court rejected. The court sanctioned the supervising attorney $2,500 as he did not accept responsibility when pressing the argument that he bore no obligation to ensure the accuracy of the citations. The associate attorney was sanctioned $1,500. The clerk was ordered to serve a copy of the order on the State Bar and both attorneys were ordered to personally report the sanctions to the State Bar.
The California Supreme Court briefly addressed the use of AI tools in a footnote in Cervantes v. Bianco (2026) __ Cal.5th __, n. 7. In preliminary briefing, respondent based a standing argument on misquoted passages from, and an incorrect recitation of the holding of, a California Supreme Court case. The law firm responsible for the error apologized and withdrew the standing arguments. The Supreme Court disapproved of the improper use of artificial intelligence tools, which contributed to the error. The Court reminded counsel of attorneys’ professional duties of competence and candor to the court and referred the matter to the State Bar of California to determine whether discipline is warranted.
Sanctions in the trial court
CCP § 128.7 applies in both the trial court and the Court of Appeal. If the court issues a sua sponte order to show cause regarding imposition of sanctions under CCP §128.7(c)(2), the court is required to provide 21 days’ notice from the date of the service of the order to show cause (OSC) re sanctions. The court may not impose monetary sanctions if, within 21 days after service of the order to show cause, the challenged paper is withdrawn or appropriately corrected. (CCP § 128.7(c)(2).)
There must be 21 days between the OSC and the hearing, plus five days for service of the OSC by mail or two days for electronic service, if applicable.
Two additional sanctions rules are available in the Court of Appeal but not in the trial court: CRC Rules 8.204(a)(1)(B) (requirements for briefs in the Court of Appeal) and 8.276 (including in the record matter not reasonably material to the appeal’s determination). The Court of Appeal does not have to provide a 21-day safe harbor if it imposes sanctions based on a violation of the California Rules of Court.
If you are facing OSC re section 128.7 sanctions, consider consulting Chapter 9, Part III, entitled Nondiscovery Sanctions, in the “bible” on civil procedure, Weil and Brown, Cal. Practice Guide: Civil Procedure Before Trial (Rutter Group 2026).
CCP § 128.7 sanctions
Generally, the court must cite a California statute, California Rule of Court or other ground for imposition of monetary sanctions.
When signing a pleading, counsel make certain certifications, to the best of counsel’s knowledge, information and belief, formed after reasonable inquiry under the circumstances.
An OSC re sanctions under CCP § 128.7 should contain the following elements: OSC re [imposition of monetary sanctions/terminating sanctions/dismissal/striking answer] [against attorney, law firm or self-represented party][against represented party, but not for meritless contentions of law] pursuant to [CCP § 128.7(c)(2)] for [specific conduct, e.g., citing hallucinated cases in motion for summary judgment in violation of CCP §128.7(b)(2)(A)] on [date of hearing 21 days from now plus 5 days for service by mail, 2 days for electronic service] at [time and place of hearing].
If the OSC does not contain all of the elements, file a written objection before the hearing on the order to show cause to preserve the issue for appeal.
The order to show cause must be issued before a voluntary dismissal or settlement of the claims made by or against the party or attorneys to be sanctioned. (CCP § 128.7(d)(2).)
Within the 21-day safe harbor period, you should withdraw or correct the target pleading. You might file a “Notice of Correction/Withdrawal [of name of pleading and date filed] Pursuant to CCP § 128.7(c)(2).” There is authority for filing an entirely new, and possibly untimely, brief that only corrects hallucinated cases, quotes or holdings. In Shayan v. Shakib (2025) 116 Cal.App.5th 619, the Court of Appeal struck the appellate brief with fabricated citations and quotations and ordered counsel to file a new brief correcting and omitting fabricated citations and quotations.
A written order imposing section 128.7 sanctions is not required, but the court must describe the conduct determined to constitute a violation and explain the basis for the sanctions imposed.
Sanctions may compensate the government for its expense in processing, reviewing and deciding the frivolous appeal. (Noland v. Land of the Free, L.P. (2025) 114 Cal.App.5th 426, 447.)
Threshold for reporting nondiscovery sanctions to the State Bar of California
The court must report nondiscovery sanctions of $1,000 or more to the State Bar. (Bus. and Prof. Code, § 6086.7, subd. (a)(3).) The appellate counterpart is California Rules of Court, Rule 10.1017.
Counsel must report nondiscovery judicial sanctions of $1,000 or more to the State Bar within thirty days of the time the attorney has knowledge of such an order. (Bus. and Prof. Code, § 6068, subd. (o)(3).)
Response to OSC and appearance at hearing
Reports to the State Bar may affect careers. Anticipate a State Bar report and aim for a warning/diversion rather than public discipline. If sanctions of $1,000 or more are imposed, the State Bar will look at all aspects of how counsel handled the matter.
The attorney who signed the target paper should also sign any written response to the OSC and appear at the hearing in person. Immediately accept responsibility, sincerely apologize, and describe steps you have taken to make sure this does not happen again. When responding, reach out to senior lawyers for help. Ask for sanctions under the $1,000 State Bar reporting threshold, e.g., $999.99.
CCP § 177.5 sanctions
If the court has issued a case management or other order that prohibits citation of fake cases, fake quotes, misrepresenting the holding of a case, and/or proffering deepfake exhibits, sanctions may be imposed for violation of a court order without good cause or substantial justification. (CCP § 177.5.)
Section 177.5 does not have safe-harbor requirements. An OSC, notice of opportunity to be heard, and a hearing are required. Monetary sanctions payable to the Court of up to $1,500 may be issued against an attorney, party or witness. For failure to comply with an order to pay sanctions, additional sanctions of up to $1,500 may be imposed. The order imposing sanctions must recite in detail the conduct or circumstances justifying the order.
Risk mitigation
Consider instituting a firmwide AI policy that specifies which tools are acceptable to use. Consider providing specialized training on how to prevent use of hallucinated materials in court filings, including training on use of detailed prompts. Make sure you and the lawyers in your firm understand the risk of hallucinations with each AI tool.
Every citation now involves risk. String cites should be a thing of the past. One or two cases on point with pinpoint citations are generally effective. Consider citing neutral, respected secondary authority such as the Rutter Group. For discovery motions, consider citing statutes only.
Develop cite-checked firmwide standard briefs or points and authorities for routine motions. File shorter, more succinct briefs.
Deepfake evidence: Photos, videos and declarations
As of the date this article was submitted, there is no published California appellate case on deepfake exhibits.
In Mendones v. Cushman & Wakefield, Inc., No. 23CV028772 (Alameda Superior Court), two self-represented plaintiffs submitted deepfake videos, screenshots of text messages, and voicemail messages in support of their motion for summary judgment.
The court issued an OSC re terminating sanctions for intentional submission of false testimony. The court found that issue and evidentiary sanctions would not serve as an adequate deterrent and issued terminating sanctions against plaintiffs for submitting fabricated evidence in support of their motion for summary judgment.
The Court stated that it has zero tolerance for deepfake evidence and found that dismissal of plaintiffs’ case would send an appropriately chilling message to litigants. The overworked and underfunded Superior Court lacked robust safeguards against future misuse. The court had lingering suspicions regarding other exhibits, but lacked the funding, time and forensic expertise to determine their authenticity. (This order is included in the DamienCharlotin.com deepfakes in court database.)
Analysis of possible deepfake evidence
The California Evidence Code provides an ample framework for evidentiary disputes regarding possible deepfake evidence.
Evidence Code section 250 defines a “writing” broadly enough to capture every photograph, video, and sound recording, so digital media are subject to the authentication rules regardless of how they are stored. No evidence is admissible except relevant evidence and except as otherwise provided, all relevant evidence is admissible. (Evid. Code, §§ 350, 351.)
The heart of the analysis is authentication. Section 1401 requires a writing to be authenticated before it is received in evidence. Section 1400 defines that task as introducing evidence sufficient to sustain a finding that the item is what its proponent claims. The Law Revision Commission Comments to Evidence Code section 1400 provide that when a judge permits a writing to be admitted in evidence, all that has been determined is that there has been a sufficient showing of the authenticity to permit the trier of fact to find that it is authentic. The judge rules on admissibility, not the weight of the evidence. The trier of fact independently determines the question of authenticity and may find that the exhibit is not authentic. The standard is a prima facie showing, not proof of genuineness.
The authenticity of a writing is a preliminary fact under section 403(a)(3), which places on the proponent the burden of producing enough evidence to support a finding of authenticity; once that threshold is met, the jury decides whether the item is genuine. Section 402 supplies the mechanism – a hearing, outside the jury’s presence, at which a deepfake challenge can be litigated before the evidence is shown to the jury. Section 405 governs the allocation of proof for related preliminary facts.
A proponent may carry its burden in a number of ways. Section 1410 provides that the enumerated methods of authentication are not exhaustive, so authenticity may be shown by a witness who perceived the event depicted, by chain-of-custody evidence, by metadata, or by expert analysis. Section 1421 permits authentication through distinctive content known only to the purported author, and section 1414 through the opposing party’s admission or its having acted on the item as authentic.
Existence and content presumptions apply to computer records and video and digital imaging. These presumptions of fact affect the burden of producing evidence, not the burden of proof.
Section 1553 presumes that a printed representation of images stored on a video or digital medium accurately represents those images, and section 1552 does the same for computer information and programs. If the opponent introduces evidence that the evidence is inaccurate or unreliable, the presumption disappears, and the proponent of the evidence has the burden of proving, by a preponderance of the evidence, that it is an accurate representation of the existence and content of what it purports to represent. (See People v. Goldsmith (2014) 59 Cal.4th 258, 269 [video and photos from red light cameras; existence and content presumptions under sections 1552 and 1553]; People v. Rekte (2015) 232 Cal.App.4th 1237, 1244-1248 [red light camera case; excellent discussion of existence and content presumptions].)
The final step in the analysis is section 352. Even where the minimal authentication threshold is arguably satisfied, the court retains discretion to exclude the item if its probative value is substantially outweighed by the danger of undue prejudice, confusion of the issues, misleading the jury, or undue consumption of time. That discretion is reinforced by the trial court’s gatekeeping duty over the expert testimony on which deepfake disputes will sometimes turn. Sections 801 and 802 empower the court to exclude expert opinion that rests on improper matter, unsupported reasoning, or speculation, a natural filter for proffered AI-detection analysis. (Sargon Enterprises, Inc. v. University of Southern California (2012) 55 Cal.4th 747, 771-772.)
Attorney deskilling
AI is not a good lawyer. It misses issues and produces hallucinations. Use AI as a thought partner. Do your own work first, then use AI. Use your brain. Develop your skills.
“Judgment and experience remain the lawyer’s own, earned over years that no tool can shortcut, and they are the very things that a client is paying for.” (Elliott v. New York Bariatric Group, LLC, No. AAN-CV-25-6066141-S (Conn. Super. Ct., Ansonia-Milford Jud. Dist., Aug. 6, 2026) at p. 10 [AI injection case].)
Conclusion
AI is transformative technology, and the legal profession should adopt it thoughtfully and intentionally. That means understanding its risks and accepting responsibility for every filing, citation and ruling that bears our names. Used wisely, with professional judgment firmly in the lead, AI can serve the legal profession without compromising its standards.
Judge Mary Ann Murphy has served on the Superior Court for 33 years, presiding over more than 400 civil jury trials. She currently sits in a trial court at the Spring Street Courthouse. She was the first judge to lead Best Practices for Mosk Central Civil judges, a role she held for nine years. Her leadership produced numerous improvements to the administration of justice, including the first asbestos working group. Elected to five terms on the Superior Court’s Executive Committee, she has also served on the Judicial Council Civil and Small Claims Committee and on statewide working groups and court committees. She is a former officer and executive board member of the California Judges Association, and she was named a 2023 Metropolitan News Person of the Year. For seven years, Judge Murphy served as an Associate Editor of Weil and Brown, Civil Procedure Before Trial. She has taught and moderated numerous programs for judges and attorneys, and she enjoys legal writing, teaching, and mentoring law students.
