Skip to main content
← predictions
55%

At least one G20 country will ban the deployment of a specific named AI model by December 31, 2026

✓ correct· 271d

Resolution

CORRECT — Italy, a G20 member, banned DeepSeek by name on January 30, 2025, meeting all resolution criteria.

The ban: Italy's data protection authority (Garante per la protezione dei dati personali) ordered DeepSeek blocked from Italian app stores and barred from processing Italian users' personal data. The order specifically named "DeepSeek" — not a generic capability class. DeepSeek was removed from both the Italian Apple App Store and Google Play Store.

Legal basis: The Garante acted under GDPR enforcement powers, issuing a legally binding order (not a recommendation or guideline). DeepSeek had failed to provide adequate information about how it processes Italian users' data, where data is stored, and what legal basis it relies on for processing.

Still in effect: As of March 2026, the ban remains active. DeepSeek has not satisfied Italy's GDPR requirements. The web version may still be accessible to some Italian users, but the official government order prohibiting deployment stands.

Additional G20 actions (not required for resolution but noted): South Korea (G20 member) also removed DeepSeek from app stores in February 2025, though this ban was lifted in April 2025 after a privacy policy update. Australia (G20 member) banned DeepSeek on all government devices in February 2025.

Sources:

  1. The Record: "Italy blocks Chinese AI tool DeepSeek" (https://therecord.media/italy-blocks-chinese-ai-tool-deepseek-over-privacy-concerns)
  2. Al Jazeera: "Which countries have banned DeepSeek and why?" (https://www.aljazeera.com/news/2025/2/6/which-countries-have-banned-deepseek-and-why)
  3. TechCrunch: "DeepSeek — the countries and agencies that have banned it" (https://techcrunch.com/2025/02/03/deepseek-the-countries-and-agencies-that-have-banned-the-ai-companys-tech/)
  4. Aitechtonic: "DeepSeek AI Banned Countries 2026" (https://aitechtonic.com/deepseek-ai-banned-countries/)

Evidence

Resolution Criteria

This prediction resolves TRUE if a G20 member country officially bans a specific AI model meeting ALL criteria:

  1. Government Action: Official law, regulation, executive order, or equivalent government decree
  2. G20 Member: Must be enacted by one of the 20 G20 countries (Argentina, Australia, Brazil, Canada, China, France, Germany, India, Indonesia, Italy, Japan, Mexico, Russia, Saudi Arabia, South Africa, South Korea, Turkey, UK, USA, EU)
  3. Specific Model: Must name a specific AI model or version (e.g., "GPT-4", "Claude-3", "Gemini Pro") - generic capability bans don't count
  4. Deployment Ban: Must prohibit deployment, operation, or use (not just training or development)
  5. Timeline: Action must be enacted by December 31, 2026
  6. Enforceability: Must be legally binding, not just recommendations or guidelines

Edge Cases:

  • Temporary emergency bans count if legally binding
  • Regional/state bans within federal countries count if they have jurisdiction
  • Bans of specific model versions (e.g., "GPT-4 Turbo") count
  • Conditional bans (e.g., "unless safety requirements met") count as bans
  • EU-wide bans count as EU being the G20 member

Exclusions:

  • General capability-based restrictions (e.g., "models above X parameters")
  • Industry-specific restrictions without model names
  • Content moderation requirements that don't ban the model itself
  • Bans on model training/development only (must affect deployment)