返回新闻
政策AI Understanding 简报

加州州长签署法案限制律师和仲裁员使用生成人工智能

州长加文·纽森 (Gavin Newsom) 签署了参议院第 574 号法案,这是一项首创的法律,禁止律师和仲裁员将法律分析委托给生成式人工智能工具,并要求在法庭文件中披露任何人工智能生成的内容。

4 min readRead the linked source
Source-provided image accompanying California governor signs bill restricting lawyer and arbitrator use of generative AI
来源参考来源记录
出版商
news.bloomberglaw.com
来源链接
news.bloomberglaw.comhttps://news.bloomberglaw.com/daily-labor-report/newsom-signs-first-of-its-kind-bill-on-lawyer-arbitrator-ai-use
来源类型
链接来源——主要来源状态尚未确定。
背景60 秒内了解这一点

从这里开始

关键术语

生成式 AI
生成文本、图像、音频、视频或代码等新内容的人工智能系统。
人工智能治理
指导人工智能如何在社会中开发和使用的政策、标准和监督机制。
引文
对模型响应中包含的源段落或文档的引用,以支持其主张。
测试一下自己人工智能道德测验

发生了什么

Governor Gavin Newsom (D) signed Senate Bill 574 on Wednesday, creating the nation’s first statutory framework that directly regulates attorneys’ and arbitrators’ use of generative artificial‑intelligence tools. The law prohibits lawyers from delegating the practice of law to AI, bans the input of confidential or personal information into AI systems, and mandates that any AI‑generated be independently verified before filing. It also requires at least one attorney to sign each court filing in their own name, unless the party is self‑representing. For arbitrators, the bill bars delegating decision‑making to AI and obliges disclosure of any AI‑derived information to the parties. Courts may impose sanctions for violations, though specific penalty structures were not detailed in the bill.

The bill, SB 574, was introduced to codify and expand upon existing California court rules that already required attorneys to verify AI‑generated . By signing the bill, Governor Newsom elevated those rules into statutory law, giving them broader enforceability.

The legislation explicitly bans attorneys from feeding confidential or personal data into tools, a safeguard against inadvertent data leaks. It also requires that any AI‑generated content used in a filing be independently checked for accuracy before submission.

Arbitrators, who often operate outside traditional court systems, are similarly constrained. They cannot rely on AI for decision‑making without full disclosure to the parties, and they are prohibited from delegating any portion of their adjudicative duties to AI.

While the bill outlines that courts may impose sanctions for violations, it does not specify exact penalty amounts or procedural steps, leaving those details to be defined by the judiciary or subsequent rule‑making.

来源详情: news.bloomberglaw.com ↗

为什么这很重要

The legislation addresses growing concerns that can produce inaccurate , “hallucinated” case law, or biased analysis, which could undermine the integrity of legal proceedings. By codifying verification and disclosure requirements, California aims to protect client confidentiality, preserve the attorney‑client privilege, and maintain public confidence in the judicial process. The bill also sets a precedent for other states and potentially the federal judiciary, signaling that regulators are willing to intervene as AI tools become embedded in professional practice. Moreover, the law could influence how law firms adopt AI, prompting the development of compliance workflows and possibly spurring the market for AI‑audit tools tailored to legal work.

Legal professionals increasingly rely on for research, drafting, and citation generation. Errors in AI output can lead to mis‑cited precedents, potentially harming client interests and eroding trust in the legal system.

The law protects client confidentiality by preventing the inadvertent transmission of sensitive data to third‑party AI providers, addressing a key privacy concern that has been highlighted in recent high‑profile cases.

By establishing a clear regulatory framework, California provides a template that other jurisdictions may follow, potentially leading to a more uniform national approach to in the legal sector.

The bill may stimulate the market for compliance‑focused AI tools, encouraging vendors to develop features that automatically verify and flag confidential data before it is sent to an AI model.

Interactive Mechanism

互动机制:它实际上是如何运作的

以交互方式探索这一发展背后的基础技术。

System Requirements:
Best ArchitecturePure RAGRecommended pattern
Hallucination RiskVery LowGrounding efficacy
Update Cost$0 (Vector sync)Ongoing maintenance
Core takeaway: Fine-tuning teaches models how to speak (form, style, syntax); RAG teaches models what to say (verifiable facts). Never use fine-tuning alone for factual memory.
交互式概念检查+10 Points
AI Ethics Quiz

Impossibility results in algorithmic fairness (e.g. Kleinberg et al., Chouldechova) show what?

接下来看什么

Key developments to monitor include: (1) how California courts implement and enforce the new sanctions regime; (2) whether other states introduce comparable statutes or adopt California’s model; (3) the response of legal‑tech vendors, who may need to add verification and audit features to their AI products; and (4) any litigation that tests the bill’s boundaries, especially regarding what constitutes “delegating” legal work to AI.

Implementation timelines: California courts will need to issue rules detailing how sanctions are applied and what procedural safeguards are required.

Legislative ripple effects: Lawmakers in other states may draft similar statutes, especially in jurisdictions with large legal markets such as New York or Texas.

Industry response: Legal‑tech companies may roll out new audit and verification modules to help firms stay compliant, creating a niche market for AI‑compliance solutions.

Litigation risk: Early cases testing the bill’s scope could clarify ambiguous terms such as “delegating” legal work or “confidential information,” shaping future enforcement.

相关指南和测验

AI 伦理人工智能代理测试你所知道的——尝试免费的人工智能测验在我们的词汇表中查找人工智能术语关注AI监管追踪器
觉得这有用吗?