Question 1

AI Hiring Tool: how does the European Union treat it?

Accepted Answer

HEAVILY REGULATED. High-risk by list position. Recruitment and selection AI: targeted job ads, CV filtering, candidate evaluation, is Annex III high-risk: conformity assessment, human oversight, bias-controlled data governance, EU database registration. Obligations apply from 2 Dec 2027 (post-Omnibus).

Question 2

AI Hiring Tool: how does the United States treat it?

Accepted Answer

HEAVILY REGULATED. The most-regulated AI use case in America, without a federal AI law. Title VII/ADA/ADEA disparate-impact liability stands (guidance withdrawn, statutes intact). NYC LL144 demands annual bias audits, Illinois AIVIA + HB 3773 cover interview and decision AI, Texas TRAIGA bans intentional discrimination, and Colorado's ADMT successor lands in 2027.

Question 3

AI Hiring Tool: how does China treat it?

Accepted Answer

MOSTLY UNREGULATED. No dedicated employment-AI law. Duties arrive sideways: PIPL Art. 24 grants explanation and refusal rights for automated decisions; public-facing platform features trigger CAC algorithm filing; Shanghai Art. 69 bars discriminatory conditions.

Question 4

LLM Chatbot: how does the European Union treat it?

Accepted Answer

TRANSPARENCY DUTY. Say you're an AI. Art. 50(1) requires users to be told they're interacting with an AI system (unless obvious): from 2 Aug 2026. The underlying model provider separately carries GPAI duties, systemic-tier if above 10²⁵ FLOPs.

Question 5

LLM Chatbot: how does the United States treat it?

Accepted Answer

LIGHT-TOUCH. A plain service bot faces no AI-specific law. Only FTC deception rules and state disclosure duties apply: California's SB 1001 bot law and Utah's on-request disclosure.

Question 6

LLM Chatbot: how does China treat it?

Accepted Answer

LICENSED: FILING REQUIRED. No filing number, no launch. Public-facing GenAI with 'public opinion attributes' needs a security assessment and CAC filing before launch; outputs are dual-labeled since 1 Sep 2025 and the filing number must be displayed. 868 services had filed by Apr 2026: this is why ChatGPT is unavailable in China.

Question 7

Medical Diagnostic AI: how does the European Union treat it?

Accepted Answer

HEAVILY REGULATED. Automatic high-risk via the device route. AI needing third-party conformity assessment under the MDR/IVDR is automatically high-risk (Art. 6(1) + Annex I): one integrated CE/Notified-Body route. AI-Act application is deferred to 2 Aug 2028; the MDR itself fully applies today.

Question 8

Medical Diagnostic AI: how does the United States treat it?

Accepted Answer

HEAVILY REGULATED. The FDA is America's most mature AI regulator. Full premarket review (510(k)/De Novo/PMA) with lifecycle and bias expectations; 1,247+ AI devices authorized, roughly 76 to 80% in radiology. PCCP guidance lets models update pre-authorized.

Question 9

Medical Diagnostic AI: how does China treat it?

Accepted Answer

LICENSED: FILING REQUIRED. NMPA device registration, typically Class III. Pre-market registration, clinical validation and lifecycle oversight; the first AI diagnostic was approved in 2020, with 50+ deep-learning devices on the market by 2023.

Question 10

Deepfake Generator: how does the European Union treat it?

Accepted Answer

TRANSPARENCY DUTY. Label it, machine-readably. Providers embed machine-readable marking (Art. 50(2)); deployers visibly disclose deepfakes (Art. 50(4)), with a relaxed disclosure mode for artistic and satirical works.

Question 11

Deepfake Generator: how does the United States treat it?

Accepted Answer

LIGHT-TOUCH (TOOL) and PROHIBITED (OUTPUTS). No law licenses the tool; liability hits the misuse. The TAKE IT DOWN Act criminalizes nonconsensual intimate deepfakes and the FTC enforces 48-hour platform takedowns ($53,088/violation); ~20 states restrict election deepfakes; the FCC killed AI-voice robocalls; California watermarking starts 2 Aug 2026.

Question 12

Deepfake Generator: how does China treat it?

Accepted Answer

HEAVILY REGULATED. The world's most detailed synthetic-media regime. Algorithm filing, security assessment, real-name verification, consent of impersonated persons, prominent labeling and GB 45438-2025 metadata; tampering is prohibited and the Criminal Law provides the teeth.

Question 13

Credit Scoring AI: how does the European Union treat it?

Accepted Answer

HEAVILY REGULATED. High-risk, plus a rights impact assessment. Annex III 5(b) makes creditworthiness evaluation of natural persons high-risk (fraud detection carved out), and deployer banks must run a fundamental-rights impact assessment (Art. 27). Applies from 2 Dec 2027; GDPR Art. 22 safeguards already apply today.

Question 14

Credit Scoring AI: how does the United States treat it?

Accepted Answer

HEAVILY REGULATED. 'The algorithm is too complex' is not a defense. ECOA/Reg B demands specific, accurate denial reasons even after the CFPB withdrew its circulars in May 2025.

Question 15

Credit Scoring AI: how does China treat it?

Accepted Answer

HEAVILY REGULATED. Credit scoring is a licensed business. A PBOC license is required; models must be explainable, traceable and filed; credit data is stored in mainland China; and PIPL Art. 24 bans 'big data price discrimination.'

Question 16

Police Live Facial Recognition: how does the European Union treat it?

Accepted Answer

PROHIBITED. Banned, with three narrow exceptions. Real-time remote biometric ID in public for law enforcement has been banned since Feb 2025: exceptions: missing/abduction victims, imminent terror or life threats, serious-crime suspects (≥4-yr, Annex II). Even then: prior judicial authorization, a FRIA, registration. Member states may ban entirely. Fines to €35M/7%.

Question 17

Police Live Facial Recognition: how does the United States treat it?

Accepted Answer

MOSTLY UNREGULATED. No federal law; a municipal patchwork. Bans in ~17+ localities (San Francisco 2019 first; Boston and Portland 2020) and state limits (Maine 2021); much of the country has none, despite documented wrongful arrests from face matches.

Question 18

Police Live Facial Recognition: how does China treat it?

Accepted Answer

HEAVILY REGULATED. Permissive for the state, strict for companies. PIPL Art. 26 permits public-security camera networks without individual consent; commercial use faces consent, DPIA, local-storage and mandatory-scan bans, plus a 100k-faces CAC filing duty: the opposite polarity of the EU.

Same AI. Three jurisdictions.

Choose the system on trial

Every verdict on the board, written out

Case 01: AI Hiring Tool

AI Hiring Tool: how does the European Union treat it?

AI Hiring Tool: how does the United States treat it?

AI Hiring Tool: how does China treat it?

Case 02: LLM Chatbot

LLM Chatbot: how does the European Union treat it?

LLM Chatbot: how does the United States treat it?

LLM Chatbot: how does China treat it?

Case 03: Medical Diagnostic AI

Medical Diagnostic AI: how does the European Union treat it?

Medical Diagnostic AI: how does the United States treat it?

Medical Diagnostic AI: how does China treat it?

Case 04: Deepfake Generator

Deepfake Generator: how does the European Union treat it?

Deepfake Generator: how does the United States treat it?

Deepfake Generator: how does China treat it?

Case 05: Credit Scoring AI

Credit Scoring AI: how does the European Union treat it?

Credit Scoring AI: how does the United States treat it?

Credit Scoring AI: how does China treat it?

Case 06: Police Live Facial Recognition

Police Live Facial Recognition: how does the European Union treat it?

Police Live Facial Recognition: how does the United States treat it?

Police Live Facial Recognition: how does China treat it?

Three patterns worth remembering

Context, not technology.

The enforcer you did not expect.

Permission before launch.

Knowing the verdict is step one. Passing the assessment is the job.