引導式測驗 · 中等 等級
律師使用 AI 工具時的客戶保密義務測驗
測試保密義務、客戶同意、提供者條款、匿名化,以及特權與律師人工智慧使用中的保密性有何不同。
問題 1 的 8
Which information may fall under a lawyer’s confidentiality duty?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- Which information may fall under a lawyer’s confidentiality duty?
- Only information labeled attorney-client privileged
- Only a client’s name and account number
- Only documents that will be submitted to court
- Information relating to the representation, even if it is not privileged or publicly filed
- A lawyer wants to paste confidential case facts into a self-learning AI tool. What does ABA Formal Opinion 512 say?
- The lawyer should obtain informed consent before inputting representation-related information when the tool may disclose it
- Consent is never needed if the tool is used for legal research
- Consent is automatically supplied by a boilerplate engagement clause
- Only the AI vendor must authorize the use of client information
- When may AI use require no client informed consent under the ABA opinion’s discussion?
- Whenever client names are removed, regardless of remaining facts
- Whenever a model says it does not train on prompts
- When the task uses no information relating to the representation, such as generic idea generation
- Whenever a lawyer signs a business account agreement
- Why might removing a client’s name fail to anonymize a prompt?
- Names are the only data providers retain
- A combination of facts, dates, locations, or unusual events may still identify the client
- Anonymization is never permitted under Rule 1.6
- AI tools can always retrieve a client identity from any fact
- What vendor information should a lawyer review before using client material?
- Only the model’s parameter count and response speed
- The provider’s marketing statements without its contractual terms
- The number of other firms that use the service
- Retention, access, training use, subprocessors, security, and deletion terms
- Does a confidential client prompt automatically receive attorney-client privilege?
- No, because privilege does not apply to electronic communications
- No; confidentiality and privilege are distinct, and a prompt is not privileged merely because a lawyer entered it
- Yes, all information typed by counsel is privileged
- Yes, if the tool has an enterprise plan
- Which practice best reduces confidentiality risk when using an approved legal AI tool?
- Minimize information, limit access, verify settings and retention, and use human review
- Paste full client files because approved tools cannot expose data
- Disable all review logs so no one can trace use
- Rely on anonymization without testing for indirect identifiers
- When can client communication about AI use be required under Model Rule 1.4?
- For every use of any software, including spell check
- Never, because Rule 1.6 covers all communication duties
- When the circumstances make use material to the representation, client expectations, or a significant decision
- Only if the client asks about AI after a filing