Quiz sprawdzający sankcje wobec AI

Odpowiedzi i wyjaśnienia pojawiają się w miarę rozwiązywania quizu.

1. Under OFAC's 50 Percent Rule, when is a company that appears on no sanctions list treated as blocked?
   • When it is based in a sanctioned country
   • When blocked persons together own 50 percent or more of it
   • When half its customers are foreign
   • When it has a similar name to a listed entity
2. What standard of liability applies to U.S. civil sanctions violations?
   • Liability only for willful violations
   • Liability only when a firm had actual knowledge
   • Liability only if a regulator issued a prior warning
   • Strict liability, even without knowledge
3. What is the purpose of below-the-line testing in sanctions screening?
   • Sampling alerts just under the threshold to see whether true matches are being suppressed
   • Reviewing only the highest-scoring alerts
   • Testing payment message formats
   • Checking staff training completion
4. Why does the guide warn about training alert-scoring models on past analyst decisions?
   • Past decisions are always correct but too few
   • Regulators ban machine learning in screening
   • True hits are rare and the labels carry past analysts' mistakes
   • Historical data is always in the wrong language
5. Why can Chinese names produce matching problems in sanctions screening?
   • Chinese names are never included on sanctions lists
   • Chinese names always have more than five parts
   • Screening systems cannot store Chinese characters
   • The same name may be romanized in pinyin, Wade-Giles or local variants
6. Which information does the guide say an alert-scoring model uses to lower the priority of a partial name match?
   • Secondary identifiers such as date of birth and nationality
   • The customer's account balance
   • The time of day the alert fired
   • The analyst's workload
7. Which of these is one of the five components in OFAC's 2019 Framework for OFAC Compliance Commitments?
   • Customer marketing
   • Risk assessment
   • Revenue forecasting
   • Product pricing
8. According to the guide, why isn't raising the matching threshold alone a good fix for false positives?
   • It makes screening slower
   • It violates the 50 Percent Rule
   • It cuts false positives but misses more true hits
   • It increases the number of alerts