引導式測驗 · 中等 等級
金融服務中的人工智慧洗錢測驗
測試人工智慧清洗定義、SEC 範例、索賠證實、揭露一致性以及金融服務投資者盡職調查。
問題 1 的 8
AI洗白描述的是什麼現象?
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。
- AI洗白描述的是什麼現象?
- Failing to publish source code for every model used
- Using an algorithm in a financial product, regardless of claims
- Having a model that performs poorly despite accurately describing it
- Overstating or misrepresenting how AI is used
- SEC在2024年對Delphia和Global Predictions的案件說明了什麼?
- Existing securities and marketing rules apply to misleading AI claims
- A firm must register every internal model as a separate adviser
- Existing securities laws do not apply when a model is proprietary
- An AI claim is exempt from antifraud rules if a vendor supplied the tool
- 企業應保留什麼來佐證其AI行銷?
- A vendor proposal without records of what reached production
- A backtest presented as proof of live deployment and monitoring
- System records, data flows, deployment evidence, and performance support
- An AI policy document without evidence that the system performed the task
- How should a prototype be described to investors?
- Describe the planned capability as live while launch is still pending
- As a prototype or research effort, not a current production feature
- Use “AI-enabled” without clarifying what the system currently does
- Report research results as current customer-facing functionality
- To distinguish an operational AI capability from a marketing label, what should a claim identify?
- Whether the system is generative or predictive without asking how it is deployed
- Whether the firm identifies the model family but not its financial use
- Whether the vendor discloses its training-data categories without describing deployment
- What task the AI performs and whether it affects advice or operations
- Which official source can an investor use to check an adviser’s registration record?
- The company’s investor-relations page as a substitute for adviser registration
- SEC/IAPD or the relevant regulator’s database
- A state corporate registry as a substitute for an adviser record
- A vendor model card instead of a regulator registration database
- Does use of AI prove better returns?
- No; a label does not show that performance improves
- Yes, if an independent vendor audited the model architecture
- Yes, if a pitch deck reports a successful historical backtest
- Yes, if the firm has filed a patent related to machine learning
- What should a compliance team compare?
- Marketing statements against vendor materials alone
- A presentation against a backtest without checking deployment
- Marketing materials with filings, deployment, and evidence
- Product language against an AI policy without performance evidence