AI Regulation
Definisi
AI regulation adalah regulasi pemerintah yang mengatur riset, deployment, dan dampak AI untuk memastikan AI aman, etis, dan bermanfaat.
Regulasi Penting
EU AI Act (2024)
- Risk-based approach:
- Unacceptable risk (dilarang): social scoring, manipulative AI
- High risk (regulasi ketat): medical, employment, law enforcement
- Limited risk (transparansi): chatbot, deepfake
- Minimal risk (tidak ada regulasi): spam filter, game AI
- General Purpose AI (GPAI) rules: transparency, copyright
- Ekstra: real-time biometric surveillance (dengan exception)
China Generative AI Rules (2023)
- AI-generated content harus “sesuai dengan nilai-nilai社会主义”
- Wajib label deepfake
- Wajib uji keamanan sebelum deployment
- Provider harus terdaftar
US Executive Order 14110 (Oktober 2023)
- Wajib lapor ke pemerintah untuk training frontier model
- Safety testing untuk dual-use foundation models
- NSF (National Science Foundation) fokus AI safety research
California SB 1047 (2024, Vetoed)
- Safety testing untuk frontier model
- Vetoed oleh Gubernur Newsom (Oktober 2024)
Internasional
- AI Safety Summit (UK, 2023, 2024)
- Bletchley Declaration (2023) — AS, UK, EU, China, India
- OECD AI Principles
- G7 Hiroshima Process
- UNESCO Recommendation on Ethics of AI (2021)
Pendekatan
- Precautionary (EU) — larang/regulasi dulu
- Innovation-first (US) — pasar dulu, regulasi kemudian
- State-led (China) — negara mengatur langsung
- Industry-led (UK, Japan) — self-regulation