6.3 AI Ethical Principles and Legal Professional Ethics

When Code Meets the Robe: The Interplay and Adaptation of AI Ethical Principles and Legal Professional Ethics

The rapid development of Artificial Intelligence (AI) has not only triggered a breathtaking wave of technological revolution globally but has also sparked profound and widespread ethical reflection and discussion. To ensure that this technology, with its infinite potential and far-reaching impact, advances healthily towards enhancing human well-being, upholding social fairness, respecting fundamental rights, and ensuring sustainable development, academia, industry, governments, and various international organizations worldwide are actively exploring, researching, and proposing various AI Ethics Principles. These principles, like lighthouses guiding the colossal ship of AI in a vast ocean, attempt to set fundamental value directions and ethical benchmarks for its development.

Simultaneously, the legal profession, an ancient and esteemed field entrusted with the core mission of upholding social justice and safeguarding citizens’ rights, possesses its own set of professional ethical norms (Legal Ethics / Rules of Professional Conduct), which have been accumulated over centuries of history, continuously refined through practice, are relatively mature in structure, and carry mandatory disciplinary force (violations can lead to warnings, fines, suspension, disbarment, or even criminal liability). In jurisdictions around the globe (e.g., the ABA Model Rules of Professional Conduct in the US, the SRA Principles in the UK, China’s Law on Lawyers and related codes), these norms serve as the “code of conduct” and “ethical baseline” that every legal professional must constantly adhere to in practice. They aim to regulate lawyers’ behavior, protect clients’ legitimate rights and interests, maintain the overall reputation and credibility of the legal profession, and ultimately serve the fairness, integrity, and efficiency of the judicial process.

When emerging, increasingly powerful AI technology is introduced into the legal practice domain—an area characterized by strict rules, significant responsibilities, and extremely high demands for accuracy and fairness—these two seemingly disparate systems of norms, the AI ethical principles representing cutting-edge technological ethical reflection and the legal professional ethics embodying centuries of industry tradition, inevitably engage in profound interplay and complex interaction. While they are highly consistent and mutually reinforcing in many core values (like fairness, responsibility, confidentiality), they may also generate inherent tensions, potential conflicts, or even pose entirely new challenges requiring reinterpretation and adaptation of traditional norms at the level of specific application scenarios and technical details.

Deeply understanding this interplay between normative systems, keenly identifying the new ethical dilemmas potentially triggered by AI applications, and proactively considering how to effectively adapt and integrate the two in concrete legal practice hold immense practical and long-term value for every legal professional aiming to responsibly, compliantly, and wisely use AI technology to enhance their work while steadfastly upholding professional integrity and industry reputation. This is not merely a matter of technology application; it fundamentally concerns how we redefine and adhere to “what it means to be an excellent, trustworthy legal professional” in the age of AI.

I. Overview of General AI Ethical Principles: Universal Value Beacons Guiding Technology Towards Good

Although specific wording, frameworks, and emphasis may vary depending on the proposing institution (e.g., government regulatory principles, tech giant self-regulatory commitments, academic ethical guidelines, or recommendations from international organizations like UNESCO), the core principles around which the global discussion on AI ethics generally revolves and gradually converges typically center on the following key value concepts. These principles provide an important, broadly applicable framework for thinking about and evaluating the ethical dimensions of AI technology at a macro level:

1. Beneficence / Human Well-being

• Core Essence: The fundamental motivation and ultimate goal of AI research, development, deployment, and application should be to serve the common good of humanity and the sustainable development of society as a whole. The value of technology lies in its ability to effectively solve important real-world problems (e.g., improving medical diagnosis, optimizing resource allocation, protecting the environment, promoting educational equity, advancing scientific knowledge), enhancing the well-being of all humankind, and should not merely pursue technological advancement for its own sake, satisfy the commercial interests of a few, or serve power ambitions. AI should be viewed as a tool serving human purposes, not an end in itself.

2. Non-maleficence / Safety, Security & Robustness

• Core Essence: Throughout the entire lifecycle of AI systems (from design, R&D, training, testing, deployment to operation, maintenance, and decommissioning), all reasonable and necessary technical, managerial, and procedural measures must be taken to proactively prevent, identify, assess, monitor, and minimize to the greatest extent possible the various forms of risks and potential harms they might cause. This includes both direct, physical harm (e.g., autonomous vehicle accidents, medical AI misdiagnosis) and indirect, non-physical harm (e.g., unfair opportunities due to algorithmic bias, mental or financial distress from privacy breaches, social trust erosion caused by disinformation). Sufficient resources must be invested to ensure the Safety (resistance to malicious attacks and unauthorized access), Security, Reliability (stable and accurate operation in expected scenarios), and Robustness (maintaining basic functionality and safety when facing unexpected inputs, environmental changes, or disturbances) of AI systems.

3. Human Autonomy

• Core Essence: The design and application of AI systems should respect and aim to enhance individual human autonomy, informed choice, and freedom of action. They should not, through opaque, deceptive, manipulative, or coercive means, undermine, interfere with, or replace human decisions made based on their own will and judgment. Especially in contexts involving individual rights, significant interests, or value choices, humans should always retain ultimate, meaningful control (Meaningful Human Control) over AI systems. AI should serve as a tool to augment human capabilities, not as a master depriving human agency.

4. Justice / Fairness

• Core Essence: The design, deployment, and outcomes of AI systems should not create or exacerbate unfair discrimination based on protected characteristics (gender, race, ethnicity, religion, age, disability, etc.) or other irrelevant factors. The Benefits derived from AI technology development (e.g., efficiency gains, service improvements) and the potential Risks and Burdens (e.g., job displacement, privacy risks, decision error risks) should be distributed as fairly and equitably as possible among members of society. High attention must be paid to actively addressing Algorithmic Bias, striving to ensure Procedural Justice & Equal Opportunity, and considering compensatory or adjusting measures for Distributive Justice & Equity when necessary.

5. Transparency / Explainability / Interpretability (XAI)

• Core Essence: The internal workings of AI systems, the main logical basis for their specific decisions or predictions, the key data features they rely on, as well as their capability boundaries, limitations, and potential risks, should, to the extent technically feasible and consistent with confidentiality and security requirements, be disclosed and explained in ways understandable to different audiences (users, affected individuals, regulators, the public) to an appropriate degree. Transparency and explainability are crucial foundations for building trust, promoting understanding, detecting and correcting errors, enabling effective accountability, and safeguarding users’ rights to information and appeal. The requirements for transparency and explainability should generally be higher in high-risk application scenarios with significant impact on individual rights or public interest (e.g., medical diagnosis, financial credit, judicial decision support, autonomous driving).

6. Responsibility / Accountability

• Core Essence: It must be possible to clearly define and effectively trace the responsible parties for the design, development, deployment, use of AI systems, and their potential behaviors and consequences (intended or unintended). This includes technical responsibility (e.g., ensuring system safety, reliability, performance), ethical responsibility (e.g., ensuring fairness, non-bias, respect for human rights), and legal responsibility (e.g., complying with regulations, bearing liability for damages caused). Robust governance frameworks and accountability mechanisms covering the entire AI lifecycle need to be established to ensure that when AI systems malfunction or cause harm, there is someone responsible (Who?), verifiable records (What went wrong & why?), established procedures (How?), and effective avenues for remedy or redress (What are the remedies?).

7. Privacy (Detailed in Section 6.2)

• Core Essence: All data activities involved throughout the AI system lifecycle (collection, storage, processing, analysis, use, sharing, transfer, destruction), especially when involving Personal Information (PII), particularly Sensitive Personal Information (biometric data, health records, financial accounts, location tracking, minor’s data, etc.), must strictly comply with all applicable data protection laws and regulations (e.g., PIPL in China, GDPR in EU) and fully implement their core principles (e.g., legality, fairness, necessity, good faith; notice-consent; purpose limitation; data minimization; storage limitation; ensuring data subject rights; and crucially—implementing adequate security measures). Protecting individual privacy must be treated as a fundamental prerequisite and core requirement for AI design and application.

8. Human-centricity & Value Alignment

• Core Essence: The development and application of AI should always place the Human at the center, with the ultimate aim being to serve human needs, enhance human capabilities, and promote human flourishing. AI system design and behavior must fully respect human dignity, fundamental rights (life, health, liberty, equality, personality rights, etc.), cognitive characteristics, emotional needs, cultural diversity, and shared societal ethical values. Furthermore, efforts are needed to address the so-called “Value Alignment” problem: how to ensure that the goals and behaviors of increasingly powerful AI systems (especially potential future AGI or ASI) remain aligned with the long-term interests and core ethical values of human society, rather than posing existential threats due to misaligned goals or loss of control. This is considered one of the most fundamental and challenging long-term issues in AI safety and ethics.

These general AI ethical principles, though varying in expression and emphasis, collectively outline the basic expectations and value orientations of the international community regarding the development of responsible, trustworthy, human-centric AI. They provide an important, broadly applicable value framework and reference point for considering and evaluating the ethical dimensions of AI technology at a macro level.

II. Core Legal Professional Ethics: Centuries-Old Standards Shaping Conduct and Baselines

Compared to the nascent, still evolving AI ethical principles, the legal profession itself possesses a set of long-established, relatively mature, more specific professional ethical norms with clear industry self-disciplinary or legally enforceable consequences (violations can lead to sanctions ranging from warnings to disbarment or even criminal liability). These norms, developed over centuries of practice and reflection, are not just behavioral guidelines and moral constraints for individual legal professionals but also the institutional bedrock indispensable for maintaining the independence, competence, integrity, and public credibility of the entire legal profession, safeguarding the core legitimate rights and interests of clients (or parties), and ultimately serving the fairness, efficiency of the judicial process, and the realization of the rule of law ideals.

While specific legal systems and codes of conduct may differ in wording, requirements, or emphasis (e.g., comparing China’s Law on Lawyers, Judges Law, Prosecutors Law, and related practice rules and ethical codes with the US ABA Model Rules of Professional Conduct and state-specific rules), the core spirit and fundamental ethical demands placed on legal professionals exhibit high commonality globally. These core requirements primarily include:

1. Competence: The Foundation of Professional Standing

• Core Requirement: Lawyers (and judges, prosecutors performing similar professional judgment duties) must possess the legal Knowledge, Skill, Thoroughness, and Preparation reasonably necessary for the representation (or handling the matter). This entails not only solid legal theory and familiarity with relevant substantive and procedural rules but also effective skills in analysis, research, communication, writing, advocacy, negotiation, etc.
• Evolving Standard: In today’s increasingly technology-driven world, the requirement of competence is evolving. For instance, Comment [8] to Rule 1.1 (Competence) of the ABA Model Rules explicitly states that maintaining competence requires keeping abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology. This strongly suggests that, in the age of AI, understanding and mastering how to safely, effectively, and compliantly use emerging technologies like AI to assist practice may be gradually becoming an integral part of maintaining professional competence.

2. Diligence: Commitment to the Client’s Matter

• Core Requirement: A lawyer must act with reasonable Diligence, Promptness, Commitment, and Dedication in representing a client’s interests, proactively safeguarding their maximum legitimate rights. This involves dedicating necessary time and effort, handling matters responsibly, responding timely to client needs, overcoming obstacles, and avoiding undue delay.

3. Confidentiality: The Bedrock of Client Trust

• Core Requirement: Lawyers owe a strict, broad, and generally lifelong duty of confidentiality regarding client secret information (typically all non-public information related to the representation) and personal privacy learned during practice. Except where disclosure is explicitly required by law (e.g., to prevent imminent serious crime) or authorized by the client’s fully informed, explicit written consent, lawyers must not, under any circumstances, reveal or improperly use such information to any third party. This is widely regarded as the most fundamental, core, and sacred of all lawyer’s professional duties.

4. Loyalty / Avoiding and Managing Conflicts of Interest

• Core Requirement: Lawyers must be completely loyal to the interests of their client, allowing no personal, other client’s, third party’s, or societal interests to compromise their loyalty and independent professional judgment. Lawyers must exercise utmost effort to avoid accepting or continuing representation in matters involving direct or potential conflicts of interest with current clients, former clients, or their own interests. Law firms need effective conflict checking systems and internal management mechanisms. When potential conflicts arise, timely full disclosure to affected parties and resolution according to rules (often requiring informed consent from all affected clients, provided the conflict is consentable) or declining/withdrawing representation is required.

5. Communication: Keeping Clients Informed and Empowered

• Core Requirement: Lawyers have a duty to maintain timely, adequate, effective, and candid communication with their clients regarding the subject matter of the representation. This includes:
  • Clearly explaining the legal context, potential risks, available options, and likely outcomes to enable the client to make informed decisions.
  • Promptly informing the client of significant developments, important documents received, and matters requiring client cooperation or decision.
  • Reasonably responding to client inquiries and requests for information.
  • If aware that a client expects conduct not permitted or unethical, the lawyer must explain the relevant legal and ethical limitations.

6. Reasonable Fees & Clear Billing Practices

• Core Requirement: Fees charged to clients must be reasonable and not illegal or clearly excessive. Reasonableness considers factors like the complexity of the matter, time and labor required, lawyer’s experience and reputation, risks involved, and customary local fees. The basis, calculation method, payment arrangements, etc., should be clearly and explicitly agreed upon in the retainer agreement and fully explained to the client to avoid misunderstanding or dispute.

7. Candor Toward the Tribunal & Respect for the Court

• Core Requirement: As an Officer of the Court, lawyers owe a high duty of Candor, Honesty, and Respect to the tribunal (court or arbitration panel) when participating in proceedings. Specifically, lawyers must:
  • Not knowingly make any false statement of fact or law to the tribunal.
  • Not offer evidence (witness testimony or documents) they know to be false. If learning previously offered evidence was false, take reasonable remedial measures (e.g., withdraw evidence, inform the court).
  • Under certain circumstances (rules vary by jurisdiction), lawyers have a duty to disclose controlling legal authority known to be directly adverse to their client’s position and not disclosed by opposing counsel, even if unfavorable.
  • Must comply with court rules and procedures, respect the authority of the judge (or arbitrator), and maintain professional decorum.

8. Maintaining the Integrity of the Profession & Promoting Justice

• Core Requirement: Lawyers are responsible not only to clients and courts but also, as members of the legal profession and key participants in the justice system, bear a broader responsibility to uphold the overall reputation of the profession, promote public confidence in the rule of law, and strive to achieve and maintain social fairness and justice. This requires lawyers to always act with integrity, honesty, and professionalism, avoiding any conduct detrimental to the profession’s image or obstructing the administration of justice.

9. Supervision of Subordinates & Non-lawyer Assistants

• Core Requirement: Partners, shareholders, or lawyers with direct managerial authority in a law firm have a duty to make reasonable efforts to ensure that the firm has effective measures giving reasonable assurance that all lawyers (especially less experienced ones) and non-lawyer assistants (paralegals, secretaries, administrative staff, IT support, etc.) in the firm conform their conduct to the rules of professional conduct. If subordinate misconduct occurs due to inadequate supervision, managers may also be held responsible.

These core professional ethical norms collectively form the “code of conduct” and “ethical compass” for legal practitioners, providing essential guidance and constraints in complex practice situations.

III. Intersections, Tensions & New Challenges: AI’s Impact on and Reshaping of Legal Ethics

When powerful, high-potential AI technology is introduced into the legal practice domain—characterized by strict rules, critical judgment, and high accountability—the emerging AI ethical principles (more macro-level advocacy and value guidance) and the established legal professional ethics (more specific behavioral constraints and baseline requirements) inevitably engage in profound interplay and complex interaction. While highly aligned and mutually reinforcing on many core values (fairness, responsibility, confidentiality), they may also generate unprecedented tensions, grey areas, or even potential conflicts at the level of specific application scenarios and technical details. AI application is posing new, sometimes extremely challenging, contemporary demands on how lawyers (and other legal professionals) understand and fulfill their traditional professional duties. We need to carefully examine these points of intersection, tension, and challenge:

1. Competence Gains New Meaning: Understanding Technology Becomes Fundamental

• AI Ethics Link: Beneficence (using tech to enhance service value), Safety & Reliability (identifying & avoiding AI risks), Responsibility & Accountability (being responsible for effective & safe tech use).
• Legal Ethics Extension & Challenge: In the context of AI increasingly permeating legal work, merely mastering traditional legal knowledge and skills might no longer suffice to meet the evolving standard of “competence.” As the ABA Model Rules comment explicitly notes, competence includes understanding the “benefits and risks associated with relevant technology.” This applies strongly to AI. This implies:
  • Basic AI Literacy is Essential: Legal professionals need at least a basic understanding of the AI tools they might use or rely on (especially LLMs)—their fundamental working principles, core functions, capability limits, inherent limitations, and typical risks (hallucinations, bias, outdated knowledge, security issues), as well as basic methods for using them safely and compliantly. Ignorance precludes effective, prudent use and risk management.
  • Mastery of Basic Usage & Evaluation Skills: Need to learn how to effectively use relevant AI tools to assist specific legal tasks (e.g., mastering basic prompt engineering for quality interaction with LLMs) and develop the ability to preliminarily assess the quality, reliability, and potential risks of AI outputs.
  • Ability to Identify & Manage AI-Related Risks: Need to be able to proactively identify potential data security risks, client privacy breach risks, algorithmic bias/discrimination risks, IP infringement risks, and related legal compliance risks in specific legal application scenarios, and understand basic mitigation measures.
  • Embrace Lifelong Learning & Adaptability: AI tech and related regulations are evolving extremely rapidly. Legal professionals need an open mindset and habit of continuous learning to actively monitor tech frontiers and regulatory updates, constantly refreshing their knowledge and skills to keep pace and maintain/enhance competence.
• Emerging Ethical Dilemmas & Discussion:
  • Could not using AI constitute “incompetence” or lack of “diligence”? An increasingly debated question. If mature AI tools can significantly enhance efficiency, reduce costs, or even achieve greater comprehensiveness and accuracy than purely manual methods for certain tasks (e.g., screening massive e-discovery data, quickly analyzing all recent case law in an area), could a lawyer who, due to ignorance or stubborn refusal, sticks to inefficient traditional methods, resulting in much longer turnaround times, higher client costs, or potentially missing critical information due to human limitations, be deemed in the future as failing to meet the reasonable standard of professional competence or diligence expected in the new era?
  • However, careless or over-reliant use of AI is clearly “incompetent”: The flip side is clearer and more dangerous. If a lawyer, with only superficial understanding of AI, unaware of its risks, and without verifying or critically examining its output, carelessly and excessively relies on immature, inaccurate, or insecure AI tools, basing client advice or court actions on flawed AI analysis, this undoubtedly constitutes a more severe breach of the core duties of competence and diligence. The key lies in finding the balance of “Prudent and Effective Use”: neither being resistant to technology’s reasonable empowerment nor blindly trusting it and abandoning professional judgment and ultimate responsibility. This requires wisdom and regulation.

2. Diligence Standard Needs Recalibration Between Efficiency and Depth

• AI Ethics Link: Beneficence (using AI efficiently for clients), Safety & Reliability (ensuring quality of AI-assisted work), Responsibility & Accountability (responsible for work quality & process), Explainability (understanding AI process/results to be accountable).
• Legal Ethics Extension & Challenge: AI tools, especially generative AI, can process vast information, generate drafts, and complete repetitive tasks at speeds unimaginable for humans. This drastically changes the “efficiency” baseline for certain aspects of traditional legal work. However, this absolutely does not mean lawyers can consequently lower the fundamental requirements for quality, depth, and rigor, or reduce their necessary core intellectual labor.
  • Verification Duty as Core Manifestation of Diligence: For any output generated by AI intended to support work or form part of a work product (research leads, case summaries, risk alerts, clause suggestions, document snippets), the lawyer bears the ultimate, non-delegable duty of review, verification, validation, and confirmation, which must be performed personally. Simply accepting, copy-pasting, or trusting unverified AI results without independent professional judgment is a direct violation of the duty of diligence, essentially an act of laziness and dereliction of duty. Diligence in the AI era is increasingly about effectively supervising and gatekeeping the quality of AI’s work.
  • Role Shift: From “Executor” to “Supervisor & Integrator” in Human-AI Collaboration: When using AI assistance, the lawyer’s role subtly but significantly shifts. For basic, information-processing tasks AI handles efficiently, the lawyer acts more like a supervisor, needing to guide and constrain AI’s “work process” through effective prompt engineering, and rigorously quality-check AI’s “output” just like supervising a (highly capable but fallible) junior human assistant. For higher-level tasks, the lawyer needs to be an integrator, organically fusing valid information, preliminary analyses, or drafts provided by AI with their own professional knowledge, experienced judgment, and strategic thinking to form the final, deep, valuable work product.
  • Efficiency Gains Should Translate to Value Deepening: The core goal of AI application should be to free up lawyers’ valuable time and mental energy from low-value, repetitive tasks that technology can handle efficiently. Lawyers should strategically reinvest this saved time into activities requiring more human wisdom, experience, creativity, and delivering higher professional value: e.g., deeper legal and business strategy thinking, more complex legal relationship analysis and argument construction, more creative dispute resolution design, and more empathetic communication and trust-building with clients. Efficiency gains must not be an excuse to reduce effort or lower quality standards. The ultimate aim should be to deepen professional service quality and client value.
• Ethical Dilemma & Discussion:
  • What is the new standard for “reasonable diligence” in the AI era? When AI can complete tasks in minutes that previously took hours or days (like legal research or document review), what constitutes “reasonable” diligence? If a lawyer insists on purely manual, inefficient methods, causing delays and high costs, could this be deemed failing the modern standard? Conversely, if a lawyer merely skims AI-generated summaries or risk reports without deep independent thought and verification, does that meet the diligence requirement? The profession needs deeper discussion and consensus.
  • Responsibility Attribution in Human-AI Collaborative Work: If a work product largely assisted by AI (e.g., a due diligence report, a complex contract) is later found to contain errors or omissions, how should responsibility be defined and allocated? Was it due to AI’s design flaws or limitations? Or was the lawyer negligent in prompt design, tool usage, or final review? Clear attribution can become more complex in increasingly blurred human-machine collaboration models, requiring clearer rules and records.

3. Duty of Confidentiality Faces Unprecedented Technical Challenges & Requires Fortification (Core points detailed extensively in Section 6.2, reiterated here for emphasis)

• AI Ethics Link: Privacy (core principle), Data Security (technical safeguard), Responsibility & Accountability (responsible for breaches).
• Fundamental Legal Ethics Requirement & Grave Risks from AI: Safeguarding state secrets, trade secrets, and personal privacy learned during practice is a mandatory professional discipline explicitly required by laws like China’s Law on Lawyers, and is arguably the most fundamental, core, and inviolable ethical duty. However, inputting any client-related (or even case-context-related) sensitive information into AI tools, especially those operated by third parties, based in the cloud, or with opaque data handling policies, poses inherent, significant, potentially catastrophic risks of data leakage and privacy violation. This includes risks during transmission, storage, processing (unauthorized access, improper collection/use by provider e.g., for model training, breaches due to platform vulnerabilities).
• Core Requirements & Practices: Legal professionals must place client information security above all else when considering AI use. This requires:
  • Establishing and enforcing the strictest internal AI use policies, prohibiting input of any confidential info into unapproved/insecure platforms.
  • Prioritizing technology solutions offering the highest level of data security and control (on-premise > private/compliant domestic cloud > rigorously vetted enterprise SaaS).
  • Employing data anonymization as a necessary risk mitigation measure (while understanding its limits).
  • Conducting the most rigorous security and compliance due diligence on third-party vendors and binding them with strong legal agreements.
  • Communicating adequately with clients about potential AI use involving their information, obtaining explicit informed consent when necessary. In summary, the duty of confidentiality is not diminished but becomes more complex and critical in the AI era, demanding higher vigilance, stronger risk awareness, and deeper technical understanding from lawyers to uphold it. Efficiency gains can never justify compromising confidentiality.

4. Loyalty & Conflicts of Interest May Acquire New Dimensions and Potential Pitfalls

• AI Ethics Link: Justice/Fairness (avoiding biased favoritism), Transparency (disclosing potential conflicts), Responsibility & Accountability (responsible for harm from conflicts).
• Legal Ethics Extension & Challenge: Lawyers owe absolute loyalty to their client’s interests and must diligently identify, avoid, and manage any direct or potential conflicts of interest that could compromise this loyalty or independent professional judgment. AI application, while seemingly neutral, might introduce new, more subtle potential conflict dimensions requiring vigilance:
  • Potential Bias or Interest Alignment of AI Tools/Vendors: Could the specific AI legal analysis tool, contract review platform, or research assistant you rely on have developers, owners, major investors, or key business partners with some undisclosed direct or indirect commercial ties or interest related to the opposing party, key witnesses, related entities, or even the judge/arbitrator in your case? Or, could the proprietary dataset used to train the AI model (e.g., transaction data from a specific industry, litigation cases from a certain sector) carry some systemic bias due to its origin or construction, making the tool’s analysis in related cases unconsciously favor one side or viewpoint? If lawyers fail to conduct necessary due diligence and prudent assessment of the tool’s background and potential biases, or fail to timely and fully disclose known potential conflict/bias risks to the client and obtain understanding, it could breach the duty of loyalty.
  • New Challenges for Internal Information Barriers in AI Systems: In large firms or organizations with extensive legal teams, strict internal information barriers (Ethical Walls) are usually necessary to manage information flow between teams potentially handling conflicting clients. If these different teams all use the same shared, centralized AI system (e.g., a firm-wide knowledge/case research platform, a unified intelligent document management/analysis system), how can we technically and managerially ensure effective, reliable isolation of sensitive information related to different cases within that AI system? How to prevent information (even patterns or knowledge learned by AI from one case) from being improperly accessed or utilized by another team with conflicting interests? This poses new technical and governance challenges.
  • Ethical Boundaries of Learning Across Client Data: A more cutting-edge and complex ethical question: If an AI model learns highly valuable, generally applicable industry patterns, risk regularities, or legal strategy insights by processing large amounts of data from Client A (even if anonymized/aggregated), and then utilizes this indirectly learned “knowledge” when providing services to another Client B in the same or similar industry, could this constitute some form of breach of the duty of loyalty owed to Client A (whose data was the source of knowledge)? Or does it create some potential conflict of interest between A and B needing management? This touches on complex boundaries between AI learning, knowledge generalization, and client data rights protection, currently lacking clear answers and requiring deep industry discussion.
• Considerations for Mitigation:
  • Include investigation into vendor background, ownership structure, major client base, and potentially model training data sources and algorithmic objectives (where feasible) as part of due diligence when selecting third-party AI tools, assessing potential bias or conflict risks.
  • For internally deployed shared AI systems, design and implement strict technical and procedural access control and information segregation mechanisms from the outset to ensure effective isolation between different cases or clients within the system.
  • In client agreements or AI usage notices, be as clear and transparent as possible about how client data will be used (including whether it might be used for internal model improvement, and how security/anonymization is ensured), obtaining informed consent.
  • Lawyers using any AI tool for analysis or decision support must always maintain professional independence and critical thinking, proactively considering and assessing potential biases or conflicts arising from the technology, data, or vendor, and not treat AI output as absolutely objective “oracles.”

5. Communication with Clients Requires Greater Transparency and Deeper Empathy

• AI Ethics Link: Human Autonomy (ensuring client informed decision-making), Transparency (disclosing AI use & risks), Human-centricity (focusing on client feelings & experience).
• Legal Ethics Extension & Challenge: Lawyers have a duty to maintain timely, adequate, effective, and candid communication with clients regarding their matters, ensuring clients can make informed decisions based on a full understanding of the situation and risks. AI application in legal services imposes new, higher demands on this traditional duty:
  • Where is the Boundary for Transparent Disclosure of AI Use?: To what extent are lawyers obligated to proactively inform clients that AI technology was used in handling their case? There are currently no uniform, clear legal or ethical rules globally. However, the prevailing responsible view suggests that appropriate, clear disclosure to the client is necessary and consistent with fiduciary duties and maintaining trust, at least when:
    • AI application materially impacts the handling method, expected outcome, service cost, or potential risks of the matter (e.g., deciding to use AI for large-scale discovery review significantly reduces cost but introduces new risks; or AI analysis becomes a key basis for critical litigation strategy).
    • A substantial portion of the work product delivered to the client was significantly assisted by AI (e.g., a complex contract draft mainly generated by AI and revised by the lawyer).
    • AI technology is used for direct interaction with the client (e.g., using an AI chatbot for initial consultation or information gathering requires clear disclosure of its AI identity).
    • When the client actively inquires about AI usage, the lawyer should provide honest, accurate answers. The manner and detail of disclosure should be tailored to the specific situation and client’s understanding, ensuring the client is not misled and understands AI’s role and potential limitations.
  • Ability to Explain AI-Assisted Analysis Results: When AI analysis results (e.g., a score from a complex risk model, cases recommended by a matching system, a win probability range from a predictive tool) are used by the lawyer as a significant basis for advising the client, explaining risks, or formulating strategy, the lawyer cannot simply “relay” the AI’s conclusion. The lawyer needs sufficient ability to explain to the client, roughly based on what main factors and through what (perhaps simplified) logical process the result was derived; how reliable the result is; what are its known major limitations, uncertainties, or potential error margins; and what role this result played in the overall decision-making (a deciding factor or one among many references?). Only then can the client truly understand the basis of the advice and make rational judgments. Lawyers cannot pass the AI “black box” directly to clients but must act as competent “translators” and “gatekeepers.”
  • Never Sacrifice Human Care & Empathy for Efficiency: AI tools can indeed improve efficiency in certain communication aspects (e.g., auto-generating meeting summaries, drafting standard progress update emails). However, legal service is fundamentally about serving “people,” especially when dealing with matters involving significant client interests, emotional turmoil, and requiring deep trust (e.g., criminal defense, divorce/family law, personal injury claims). In these moments, the empathetic listening, emotional understanding and support, and direct, candid, warm interpersonal communication provided by human lawyers hold a value that no cold technology can replace. Over-reliance on AI for client communication, particularly at critical moments demanding human connection and care, can severely damage the client experience, erode the hard-earned attorney-client relationship, and even make clients feel “objectified” or neglected. Lawyers need wisdom and emotional intelligence to strike the delicate balance between leveraging technology for efficiency and maintaining necessary human interaction and emotional care.

6. Principle of Reasonable Fees Faces Re-evaluation Amidst AI-Driven Efficiency Gains

• AI Ethics Link: Justice/Fairness (fees reflect true value & effort), Transparency (clearly explaining fee basis to clients).
• Legal Ethics Extension & Challenge: Fees charged by lawyers must be reasonable and not illegal or clearly excessive. The rationale behind the traditional Billable Hours model assumes fees are proportional to the professionally valuable time lawyers invest. However, when AI technology can dramatically, even exponentially, increase efficiency for certain traditionally time-consuming legal tasks (e.g., reviewing vast documents in minutes vs. days; quickly generating drafts for standard documents that took hours), if lawyers still strictly bill clients at high hourly rates based on the nominal time spent on tasks largely completed by AI, the reasonableness foundation of this billing method might be shaken. The final fee might become disproportionate to the actual human intellectual labor and professional judgment value contributed, potentially leading to client challenges or even violating ethical rules on reasonable fees.
• Responding Trends & Future Directions: The AI efficiency revolution is powerfully pushing the legal services industry to rethink and reform its traditional billing models. Alternative Fee Arrangements (AFAs) are likely to become increasingly common, such as:
  • Fixed Fees / Flat Fees: For services with relatively predictable outcomes and standardized processes (incorporation, trademark filing, standard contract drafting), agree on a total fixed fee.
  • Value-based Billing: Fees based more on the actual value created or goals achieved for the client (successful deal closing, winning a key lawsuit, avoiding major losses), rather than just time spent.
  • Tiered Fees / Capped Fees: Setting different rates for different phases of work or a total fee cap.
  • Contingency Fees / Success Fees: Where permitted by law and ethics rules (e.g., certain damages claims), fees linked to the successful outcome or recovery amount.
  • Hybrid Models: Combining elements of the above. Regardless of the model used, in the age of AI, clearer, more transparent communication with clients regarding the specific basis for fees, calculation methods, and the potential role of AI in service delivery and its impact on efficiency/cost will become more crucial than ever. This helps manage client expectations and build fair, trusting relationships.

7. Duty of Candor Toward the Tribunal Requires New Considerations Regarding AI-Generated Content

• AI Ethics Link: Responsibility & Accountability (responsible for court submissions), Transparency (disclosing AI use when necessary), Fairness (ensuring AI isn’t used to mislead court).
• Legal Ethics Extension & Challenge: As Officers of the Court, lawyers owe absolute, non-delegable duties of Candor, Honesty, and Respect to the tribunal. They must not provide false information or misleading statements. Generative AI application introduces new scenarios requiring careful handling regarding this duty:
  • Disclosure Duty & Risks for Citing AI-Generated Content: If a lawyer plans to submit legal documents (pleadings, briefs), written arguments, or even analytical reports as part of evidence to court, and significant, substantive portions were assisted by generative AI (like LLMs) (even if reviewed and revised by the lawyer), is the lawyer obligated, and to what extent, to proactively disclose the use of AI technology in preparation? Currently, there are no uniform, clear legal rules or widely accepted industry standards globally. Practices and expectations might vary significantly across jurisdictions, courts, or even individual judges. Some courts have begun issuing local practice guidelines or case-specific orders possibly requiring lawyers to make some form of declaration, assurance, or certification when using AI (e.g., confirming all cited cases and authorities were manually verified and not fabricated by AI). As AI use becomes more common, more uniform and clearer rules are likely to emerge. Until then, lawyers should uphold the highest standards of integrity and prudence. Given the inherent risks of “hallucinations” and factual errors in AI (especially LLMs), and the paramount importance of maintaining the court’s trust in the reliability of lawyers’ submissions, a more responsible approach, consistent with the spirit of candor, might be: if AI’s contribution goes beyond mere formatting or language assistance to involve substantive content generation or analysis, make appropriate, cautious disclosure (e.g., in footnotes or appendix explaining AI’s auxiliary role and human verification conducted), and explicitly state that the submitting lawyer bears full responsibility for the accuracy and legal validity of the final content. This might be safer and better preserve trust. Under no circumstances should AI-generated content be presented directly as one’s own independent work product.
  • Special Considerations for Submitting AI-Generated or Processed Evidence: (Discussed in Sections 5.6 & 6.6) When parties seek to submit evidence generated by AI (e.g., 3D scene reconstruction from photos) or substantially processed by AI (e.g., significantly enhanced blurry surveillance video, AI-transcribed court audio claimed to be verified), the submitting lawyer has a duty to:
    • Fully and clearly explain to the court the generation or processing method, the specific AI technology and algorithms used, the reliability and limitations of the technology, and potential error margins or distortion risks.
    • Be prepared to provide relevant technical details and supporting materials (possibly requiring AI expert assistance) to withstand potential challenges from the opposing party regarding the evidence’s authenticity, accuracy, reliability, or potential manipulation.
    • Ensure the entire process complies with relevant evidence rules regarding Best Evidence, Authentication, and admissibility of scientific evidence (like Daubert/Frye standards in the US).

  Zero Tolerance: Never Submit Hallucinated Content to Court!

  This should be considered the absolute bottom line of a lawyer’s duty of candor to the tribunal! Submitting any document containing fabricated cases, non-existent statutes, false evidence citations, or any other untrue information generated by AI is absolutely unforgivable, a severe violation of professional ethics and litigation rules. It can lead not only to direct case failure but also subject the lawyer to extremely severe court sanctions (including heavy fines, public reprimand, referral to bar) and potentially end their career. When using AI for research or writing, 120% verification of all facts and legal authorities is the only way to avoid this disaster.

8. Duty of Supervision Extends to Guiding and Overseeing Subordinates’ AI Use

• AI Ethics Link: Responsibility & Accountability (manager responsible for subordinate conduct), Safety (ensure safe use by subordinates), Fairness (prevent misuse leading to discrimination).
• Legal Ethics Extension & Challenge: Partners, shareholders, or senior lawyers with direct managerial or supervisory responsibilities in a law firm or legal department, their ethical Duty of Supervision extends beyond traditional work content and professional conduct of junior lawyers, trainees, and non-lawyer assistants (paralegals, secretaries, IT support). It now explicitly includes supervising how they compliantly, effectively, prudently, and responsibly use artificial intelligence tools to assist their work.
• Specific Supervisory Duties: This implies managers themselves need sufficient understanding of basic AI concepts, potential risks (especially confidentiality, hallucination, bias), and relevant internal policies to be able to:
  • Provide clear guidance and training to subordinates on proper AI tool usage methods, risk prevention measures, and ethical boundaries.
  • Reasonably assign tasks involving AI assistance, ensuring subordinates have the necessary competence and prudence for those tasks.
  • Establish effective review and checking mechanisms for substantive review and quality control of work products completed with AI assistance by subordinates (especially those intended for external submission or client impact), not just formal sign-offs.
  • Ensure subordinates strictly adhere to the organization’s AI use policy and all relevant laws, regulations, and ethical norms, especially regarding client confidentiality.
  • Promptly correct and guide subordinates if improper or risky AI usage practices are observed. If serious violations or significant losses occur due to inadequate supervision or negligence by managers regarding subordinates’ AI use, the managers themselves might face corresponding managerial or even legal liability.

IV. Seeking Harmony, Integration & Foresight: Exploring Paths to Responsible AI in Legal Application

Faced with the profound changes AI brings to legal practice, and the complex interactions, potential tensions, and novel challenges arising between AI and traditional legal professional ethics, legal professionals and the entire industry need to proactively seek a path of reconciliation, integration, and foresight. We need to explore how to fully leverage AI’s immense potential while steadfastly safeguarding the core values and ethical bottom line of the legal profession, ultimately achieving Responsible AI in Law. This likely requires collective effort in several areas:

• Embrace Lifelong Learning, Internalize AI Literacy as Core Competence: Legal professionals need to overcome fear or resistance towards new technology and view learning and understanding AI (at least applications relevant to their work) as an ongoing, necessary self-improvement task. Actively learn AI’s basic principles, key technologies (LLMs, RAG, Prompt Engineering), main applications, capability limits, typical risks, and related legal/ethical norms. Gradually internalize basic AI Literacy as a core professional competency for the new era, alongside legal research, writing, and advocacy.
• Uphold Human-centric Principles, Ensure Technology Serves Law’s Core Goals: Always remember that AI is merely a tool, a means, not an end in itself. Technology development and application must always be centered on serving the fundamental interests, rights, and well-being of people (clients, parties, the public). Technology adoption should aim to better achieve the core goals of law—e.g., enhancing efficiency and accessibility of justice, ensuring procedural fairness and transparency, promoting correct understanding and application of legal rules, maintaining social harmony and stability—and never sacrifice these fundamental values for the sake of technological coolness or extreme efficiency.
• Strengthen Professional Judgment & Critical Thinking as “Firewalls”: Precisely when AI offers seemingly powerful, convenient information processing and analysis assistance, we need to strengthen and cherish the core human capabilities AI lacks: independent critical thinking; prudent judgment based on deep legal knowledge and rich practical experience; profound insight into complex situations and subtle human nature; capacity for creative problem-solving and value balancing; and the ultimate assumption of ethical and legal responsibility. These abilities act as “firewalls,” helping us filter out noise, errors, and risks from AI output, ensuring technology application stays on the right track.
• Actively Participate in Rulemaking & Standard Setting: Legal professionals are not just users of AI technology but should also be shapers of relevant rules. Leverage professional expertise, practical experience, and deep understanding of rule-of-law principles to actively participate in discussions, formulation, and refinement of ethical guidelines, best practice guides, industry self-regulation norms, and even related laws, regulations, and judicial policies concerning AI application in the legal field. Contribute constructive insights and rationality to guide the healthy development of this transformative technology within reasonable boundaries.
• Maintain Transparency, Responsibility & Continuous Improvement in Practice:
  • Internal Transparency: Foster open communication within teams and organizations about AI tool selection, evaluation, usage, encountered problems, and lessons learned, encouraging knowledge sharing and mutual learning.
  • External Prudent Transparency: Communicate necessarily, honestly, and responsibly with clients, courts, or regulators about AI usage based on specific circumstances and relevant rules, building and maintaining trust.
  • Continuous Improvement: Treat AI application as a dynamic process requiring constant learning, reflection, evaluation, and improvement. Actively collect feedback, courageously admit and correct errors, and continuously optimize application strategies and risk management measures based on evolving technology and context.
• Prioritize Risk Prevention Strategically, Protect Core Values First: When making decisions about AI applications (selecting tools, designing processes, evaluating outcomes), always place risk prevention (especially data security, client confidentiality, accuracy assurance, compliance) at a strategic priority. When potential trade-offs arise between efficiency gains and risk control, unhesitatingly prioritize protecting clients’ core interests, upholding professional ethical baselines, and defending judicial fairness. These are the non-negotiable foundations of the legal industry; no technological application can come at their expense.

Conclusion: Ethics Lead, Norms Guide, Human-Machine Wisdom Dances for Sustainable Progress

Emerging AI ethical principles, full of potential, provide important frameworks and guidance for contemplating the macro-level societal values and ethical directions AI development should follow. Time-honored, deeply rooted legal professional ethics, on the other hand, offer clear behavioral standards, specific practical requirements, and inviolable baseline constraints for legal professionals acting responsibly in concrete practice situations.

In the context of AI increasingly integrating into legal practice, how to effectively combine, cross-reference, and dialectically understand these two normative systems, and how to continuously adapt, apply, and reflect upon them in evolving practice, is a core challenge facing every legal professional and organization.

This requires us to possess a cross-disciplinary perspective, understanding both technology’s potential/limits and law’s principles/spirit; to maintain lifelong learning capabilities, keeping pace with rapid iterations in tech and regulation; to uphold a prudent practical attitude, mindful of risks while exploring, adhering to baselines while innovating; to embrace a spirit of open exchange, sharing experiences and building consensus within the profession and across fields; and most importantly, to have an unwavering commitment to safeguarding the rule of law, professional values, and ethical responsibilities.

Only in this way can AI truly become a beneficial, reliable, trustworthy force promoting legal service quality, advancing judicial efficiency and fairness, and contributing positively to the construction of a rule-of-law society, rather than degenerating into an unmanageable “Pandora’s Box” bringing new ethical dilemmas, legal risks, and social divides. Responsible AI application must begin with profound ethical guidance, proceed under the strict escort of norms, and ultimately be perfected through the harmonious dance of human wisdom and machine intelligence. Only then can the legal profession navigate the historic waves of AI steadily and reach far.