Skip to content
AI Encyclopedia for Legal Professionals

10.5 Frequently Asked Questions (FAQ)

Section titled “Clarifying Doubts: Frequently Asked Questions (FAQ) for Legal Professionals on AI”

When exploring the increasingly intertwined fields of artificial intelligence (AI) and law, legal professionals of all experience levels may encounter various questions, confusions, and even concerns. This section aims to compile some of the most frequently asked questions encountered during learning and practice and provide concise, to-the-point answers. Hopefully, this will help clarify ambiguous concepts, alleviate unnecessary doubts, provide a clearer understanding of AI technology, enable more confident planning of learning paths, and encourage more prudent consideration of its application in legal work.

Q1: I have a purely legal background with no programming or technical foundation. Will learning AI be too difficult for me? Do I absolutely need to learn coding?

A: There’s absolutely no need to be overly concerned, nor is it mandatory to learn programming! For legal professionals, the core objective of learning AI is not to become a technical expert capable of building algorithmic models from scratch, but rather to achieve the following key levels of understanding and application ability:

  1. Understand Basic Concepts & Principles: Be able to understand discussions about AI (especially LLMs, generative AI), grasp what it is and isn’t, and know roughly how it works.
  2. Recognize Capability Boundaries & Risks: Clearly understand AI’s strengths (e.g., speed and scale of information processing) and its inherent limitations and potential risks (e.g., hallucinations, bias, confidentiality issues).
  3. Master Core Application Skills: Learn how to effectively use existing, mature AI tools to assist your legal work, especially mastering the key skill for efficient and safe interaction with generative AI like LLMs—Prompt Engineering.
  4. Develop Ethical & Compliance Awareness: Understand the ethical challenges and legal compliance requirements associated with AI applications and be able to consider them in practice.

This resource is designed precisely to help legal professionals like you understand these core concepts in relatively plain language (especially Parts 1, 2, 6, 7, 9). Mastering Prompt Engineering (see Part 4)—learning how to “ask the right questions” and “give the right instructions”—is far more direct, important, and practical for leveraging AI’s value than learning a specific programming language.

Of course, if you have a strong personal interest in technology and time permits, understanding some Python programming basics or fundamental data analysis concepts would certainly be beneficial for deeper AI understanding or participating in more complex AI projects (like collaborating with tech teams on custom models), but it is by no means a prerequisite for legal professionals to apply AI. The key is maintaining an open mindset and enthusiasm for continuous learning.

Q2: Will artificial intelligence really replace lawyers or in-house counsel? Should I be very anxious about this?

A: The prevailing view in the industry and foreseeable trends suggest: AI will not completely replace the core functions of lawyers or legal counsel, but it will inevitably and profoundly change the profession.

  • AI’s Limitations: Current and near-future AI (all narrow AI) excels at standardized, repetitive, pattern-based tasks but still falls far short compared to experienced human professionals in core legal work requiring deep legal reasoning, complex strategic judgment, creative problem-solving, profound ethical consideration, building interpersonal trust, empathetic communication, and handling high uncertainty and ambiguity. These are precisely where the core value of lawyers and legal counsel lies.
  • AI as a Tool, Accelerator & Amplifier: A more accurate understanding is that AI will become an extremely powerful auxiliary tool for legal professionals, freeing them from vast amounts of tedious work, amplifying their professional capabilities, accelerating their efficiency, and enabling them to focus more on providing higher-value services.
  • Shift in Competitive Landscape: The real challenge likely comes not from AI itself, but from peers who master and effectively utilize AI technologies. In the future, “lawyers who use AI” will likely have a significant competitive advantage over “peers who don’t or refuse to use AI” in terms of efficiency, service quality, cost control, and even business development.
  • Coping Strategy: Embrace, Don’t Fret: Therefore, instead of succumbing to anxiety about “being replaced,” adopt a positive, proactive attitude:
    1. Strive to Learn and Adapt: View AI as an opportunity to enhance your own capabilities and broaden career horizons.
    2. Focus on Core Value: Reflect on and strengthen those uniquely human advantages that AI struggles to replace (strategic thinking, communication, negotiation, ethical judgment, client relationships, etc.).
    3. Master Human-AI Collaboration: Learn how to work effectively in synergy with AI, making it your trustworthy (yet supervised) “intelligent assistant.”

Q3: Is the legal information or advice generated by Large Language Models (LLMs) like ChatGPT or DeepSeek reliable? Can I directly copy and paste it?

A: Absolutely unreliable! Absolutely must not be used directly! This is an extremely important red line issue that must be emphasized repeatedly.

  • Core Limitations of LLMs:
    • Hallucination: LLMs generate text based on probabilistic patterns, not fact-checking. They are highly prone to confidently fabricating completely non-existent legal cases, incorrect statute numbers, distorted interpretations of legal concepts, or even fictional expert opinions. (See Sections 2.8, 6.6)
    • Outdated Knowledge: LLM knowledge comes from their training data, which has a specific cutoff date. They cannot reflect the latest amendments to laws/regulations, judicial interpretations, or recent case law.
    • Lack of True Legal Reasoning Ability: LLM “reasoning” is essentially pattern-based association, not rigorous legal logical deduction. They struggle to accurately grasp legal nuances, application prerequisites, value orientations, and complex contexts.
  • Immense Risks of Direct Use: Directly using any unverified legal information or seemingly “advice” generated by LLMs in your legal research reports, drafted documents, client consultations, or court filings carries extremely high, unacceptable professional risks. This can lead not only to serious legal errors, lost cases, client damages, but also directly violate lawyers’ duties of diligence and candor to the tribunal, potentially resulting in severe legal liability or disciplinary sanctions. (There are real cases internationally where lawyers faced penalties for this; heed the warning!)
  • Correct Positioning & Usage: LLM outputs can at most be considered:
    • Preliminary, unverified sources of reference information.
    • Tools for brainstorming or broadening perspectives.
    • Sources of raw material for generating initial drafts of non-core, standardized text.
  • Mandatory Human Verification & Substantive Revision: Any LLM output intended for use must undergo extremely rigorous, independent, multi-source cross-verification, factual validation, and legal logic review by qualified and experienced legal professionals. It typically requires substantial modification, supplementation, and refinement before any part can be prudently considered for incorporation.

Q4: Is it safe to use third-party AI tools (especially cloud services) to process clients’ sensitive information? How should I protect client confidentiality?

A: Using AI tools to process sensitive client information carries significant security and privacy risks and demands the highest level of caution and the strictest protective measures!

  • Core Risk Points: Data might be transmitted to external servers, potentially used by vendors for model training, leaked due to platform security vulnerabilities, or violate cross-border data transfer regulations, etc. (See Sections 6.1, 6.2, 7.4)
  • Key Practices for Protecting Client Confidentiality:
    1. Prohibit Input into Public/Free Platforms: Absolutely forbid inputting any identifiable client information, specific case details, trade secrets, or any content protected by confidentiality/privilege into any unapproved, publicly available, or free AI platforms (like personal ChatGPT accounts, web-based translation/transcription tools).
    2. Prioritize Secure, Controllable Solutions: For sensitive data, the first choice is fully local deployment (e.g., running open-source models on internal servers); second choice is enterprise private cloud or dedicated instances/tenants on public cloud offering data isolation, strong encryption, and clear commitments against data use for training; third choice is enterprise-grade SaaS services that have passed rigorous security reviews and have robust DPAs/NDAs.
    3. Strictly Review Vendor Security & Privacy Policies: Carefully read and understand the provider’s terms regarding data processing location, storage duration, access controls, encryption measures, data usage purposes (especially for training), security certifications, and incident response/notification mechanisms. Do not accept vague or unfavorable terms regarding client confidentiality.
    4. Maximize Data Anonymization/Masking: Before input, anonymize or mask data to the greatest extent possible without significantly hindering the task, removing all unnecessary sensitive information.
    5. Obtain Client Informed Consent When Necessary: For substantial processing of important sensitive client information with AI, transparently inform the client about risks and protective measures, and obtain their explicit written consent.
    6. Strictly Adhere to Internal AI Use Policies: Ensure all actions comply with your organization’s internal rules on AI use and data security.

Protecting client confidentiality is a lawyer’s paramount professional duty, allowing absolutely no compromise. Adopt the most conservative and cautious approach always.

Q5: I often hear about “Prompt Engineering.” What exactly is it, and why is it important for legal professionals?

A:

  • Prompt: It’s the instruction, question, request, or contextual information you input into an AI (especially LLMs like ChatGPT).
  • Prompt Engineering: It’s the art and science of designing, constructing, optimizing, and iterating on these “prompts.” The goal is to more effectively guide the AI to accurately understand your intent and generate the desired high-quality, relevant, and useful output.
  • Importance: LLM output quality is critically dependent on input prompt quality. A good prompt acts like a “GPS” for the AI, guiding it to perform optimally and reducing errors/hallucinations. A poor prompt is like asking the AI to “grope an elephant in the dark,” likely yielding irrelevant or worthless results.
  • Specific Importance for Legal Professionals:
    • Legal tasks often require high precision, logical rigor, and specific formatting. Only well-crafted prompts can guide AI to meet these demands better.
    • Mastering prompt engineering can help mitigate AI risks like hallucination and bias to some extent (e.g., by providing context, setting constraints).
    • It transforms LLMs from simple Q&A “toys” into genuinely practical productivity tools capable of deeply assisting complex tasks like legal research, document drafting, contract review. Prompt engineering is the “golden key” for legal professionals to harness LLM capabilities and unlock their value. (See Part 4 for details)

Q6: Who owns the copyright to the contract diagram I generated using an AI tool (like Midjourney or Stable Diffusion), or the analysis report I drafted with ChatGPT’s assistance?

A: This is a very complex legal question with no definitive global consensus yet. The current prevailing views and emerging practice trends are roughly as follows (See Section 7.3):

  • AI Itself Cannot Be an Author: Under current copyright law, authors are generally required to be human. Therefore, content purely generated by AI automatically, lacking human original contribution, is likely ineligible for copyright protection and may fall into the public domain.
  • Human User’s Authorship Depends on “Original Contribution”: If, in using the AI tool, you did more than just give a simple command—e.g., through complex prompt design, multiple rounds of iterative refinement, creative selection/arrangement of AI-generated elements, and significant manual post-editing—such that the final work sufficiently reflects your personal, unique intellectual choices and expression, then you might be recognized as the author of the final work (or at least your contributions) and hold copyright. However, the standard for “sufficient contribution” is currently very vague and likely requires case-by-case determination.
  • Derivative Work Infringement Risk: Beware that even if you hold copyright in the AIGC, the content itself might potentially infringe the copyright of underlying works used in the AI’s training data. This is a major focus of ongoing litigation.
  • Service Terms Matter: Different AI tool providers may have varying terms of service regarding IP ownership and usage rights for generated content (e.g., some grant commercial use rights to users but retain certain rights themselves). Always carefully read and understand the specific terms of the tool you use.

Practical Advice: If planning significant commercial use of AIGC, strongly advise consulting with specialized IP lawyers to assess risks and strategies. Keep detailed records of your human creative contributions during the process.

Q7: Will artificial intelligence exacerbate discrimination issues in the legal field? How should we deal with algorithmic bias?

A: Yes, this is a very real and serious risk! AI systems (especially ML-based ones) can unfairly and systematically treat or impact specific groups in areas like hiring, credit scoring, risk assessment, even judicial assistance, by learning from historical societal biases present in training data or due to flaws in algorithm design or neglect of fairness. This can replicate, perpetuate, or even amplify social injustices.

Addressing algorithmic bias requires a comprehensive approach throughout the AI lifecycle:

  1. Improve Data Quality & Representativeness: Strive to collect more diverse, representative training data and apply cleaning and de-biasing techniques.
  2. Conduct Rigorous Fairness Audits: Before and after deploying AI systems, use multiple fairness metrics to rigorously test for bias against protected groups.
  3. Employ Bias Mitigation Techniques: Use technical methods during data processing, model training, or post-processing to try to reduce identified biases (but be mindful of potential fairness-accuracy trade-offs).
  4. Maintain Meaningful Human Oversight & Intervention: Never allow AI to make fully automated high-risk decisions impacting individual rights. Mandate strong human review, intervention, and final decision-making.
  5. Establish Transparency & Grievance Mechanisms: Increase transparency around algorithmic decision-making where possible, and provide effective channels for individuals to appeal and seek redress for potentially unfair outcomes.
  6. Comply with Anti-Discrimination Laws: Ensure all AI applications strictly adhere to relevant anti-discrimination legal requirements. (See Sections 6.1, 6.4, 7.5)

Q8: There are so many legal AI tools on the market. How should our law firm/legal department choose? Are there simple criteria?

A: There are no simple “one-size-fits-all” criteria; choosing AI tools requires a systematic, multi-dimensional evaluation process. Here are the core steps and considerations:

  1. First, Clearly Define Your Core Needs & Goals: What specific problem are you trying to solve? What concrete value do you expect AI to bring? (See Section 9.2, Step 1)
  2. Conduct Market Research & Initial Screening: Identify potential solutions that could meet your core needs.
  3. Perform Rigorous Pilot Testing & Performance Validation: “The proof of the pudding is in the eating.” Must test candidate tools with real (anonymized) data and scenarios to assess actual performance, accuracy, and usability, comparing against manual methods or alternatives.
  4. Prioritize Data Security & Privacy Compliance Above All Else: Conduct extremely rigorous scrutiny to ensure the chosen tool fully meets your confidentiality requirements and all applicable laws/regulations. (See FAQ Q4 & Sections 6.2, 7.4)
  5. Assess Cost-Effectiveness & Total Cost of Ownership (TCO): Consider all costs (software, hardware, integration, training, maintenance, usage fees) versus expected value.
  6. Evaluate Integration Capability with Existing Workflows: Can the tool integrate smoothly with your current systems and ways of working?
  7. Assess Vendor Reliability & Long-Term Support: Choose reputable vendors with proven tech, stable service, and professional support.
  8. Finally, Confirm Ethical & Broader Compliance: Revisit concerns about bias, transparency, IP rights, etc.

Remember: There’s no “best” AI tool, only the one “most suitable” for your specific needs, risk tolerance, and resource situation. (See Sections 3.4, 5.7)

Q9: Deepfake technology poses a huge threat to the authenticity of legal evidence. How should we respond?

A: Deepfakes indeed present a fundamental challenge to authenticating audio-visual evidence. Response strategies need to be multi-pronged:

  1. Increase Awareness & Skepticism: Legal professionals must recognize the existence and risks of deepfakes, no longer easily trusting seemingly real audio/video evidence.
  2. Strengthen Technical Detection Capabilities: Stay informed about and (when necessary) utilize specialized deepfake detection technologies (commercial services or forensic labs) for technical analysis of suspicious media. But understand detection tech has limits and isn’t foolproof.
  3. Rigorously Scrutinize Evidence Source & Chain of Custody: Apply stricter scrutiny to the original source, legality of acquisition, and integrity of the chain of custody for audio/video evidence from collection to court submission.
  4. Emphasize Corroboration with Multi-Source Evidence: Never rely solely on potentially fake audio/video evidence. Place high importance on comparing and corroborating it with other types of independent evidence (documents, physical evidence, other witness testimony).
  5. Rely on Expert Forensics & Court Examination: In contested situations, digital forensics experts or deepfake specialists may be needed to provide expert opinions on authenticity, subject to court examination and cross-examination.
  6. Promote Adaptation & Development of Evidence Rules: The legal community needs to collectively consider and promote updates to evidence rules to better address challenges from new tech like deepfakes, e.g., potentially higher standards for authenticating digital audio/video? How to allocate burden of proof? (See Sections 6.6, 8.2)

Q10: I want to start systematically learning about AI and law. Do you have recommended learning paths or resources?

A: Excellent initiative! Here’s a suggested learning path and types of resources:

  • Step 1: Reshape Mindset, Embrace Learning: Stay curious, overcome fear, recognize this as essential future literacy.
  • Step 2: Build Foundational Cognitive Framework:
    • Read this Resource (Legal AI Encyclopedia): Especially Part 1 (Intro & Basics), Part 2 (Core Tech Overview), Part 6 (Risk, Ethics, Governance) for core concepts, capabilities, key risks.
    • Read Introductory Books: (See Section 10.2 Book Recommendations)
  • Step 3: Focus on Core Skill - Prompt Engineering:
    • Study Part 4 of this resource for principles, techniques, legal scenario templates.
    • Extensive Hands-on Practice: Safely try mainstream LLM tools, repeatedly practice, test, refine prompts. Most crucial step.
  • Step 4: Understand Application Scenarios & Tools:
    • Read Part 5 & Sections 3/10.2: Learn about AI applications in various legal practice areas and mainstream tools/platforms.
  • Step 5: Focus on Risks, Ethics & Compliance:
    • Study Parts 6, 7, 8, 9: Deep dive into security, privacy, bias, IP, liability, professional ethics.
  • Step 6: Leverage Diverse Resources for Continuous Learning:
    • Online Courses (MOOCs): Coursera, edX, etc. (See Section 10.2).
    • Professional Media & Blogs: Reputable sources for Legal Tech, AI Ethics (See Section 10.2).
    • Academic Literature & Reports: arXiv preprints, journals, think tank reports (See Section 10.2).
    • Official Documentation: Best resource for specific AI tools you use.
    • Communities & Conferences: Engage online/offline with peers/experts (See Section 10.2).
  • Step 7: Practice, Reflect, Share: Apply learning to real (safe) work, document lessons, share insights with colleagues.

The most important thing is to start and maintain the enthusiasm and habit of continuous learning. Hopefully, this FAQ provides initial guidance. For more detailed and in-depth content, please refer to the corresponding sections of this encyclopedia.