AI, Surveillance, and the Fight for Digital Sovereignty | Near's Illia Polosukhin

— AI export controls & internet precedent: The US government's ban on Anthropic's Claude 3.5 Fable sets a dangerous precedent for restricting internet services by country, potentially fracturing global access to critical infrastructure and technology.

— Nationalization risk for AI labs: As governments recognize AI's power (comparable to nuclear weapons), they will likely move to nationalize leading AI companies to maintain control, forcing KYC requirements and full data surveillance on users.

— Decentralized AI as the countermeasure: Near is building a sovereign AI stack—confidential inference, Ironclaw agents, and private smart contracts—allowing users to access powerful models without KYC, data leakage, or government surveillance.

— User-owned alignment vs. corporate alignment: General alignment is a myth; real alignment means systems work exclusively for the individual user or specific company, not broad society. Blockchain tokenization solves OpenAI's capital formation problem and ensures user-company alignment.

— Agent marketplace & confidential compute: Specialized agents can be hired on-chain with verifiable privacy guarantees, enabling businesses to delegate sensitive work (CRM access, email, financial data) without data breach risk.

— Compute as the real bottleneck: Hyperscalers control most GPU supply through 2026, but cost arbitrage and model routing make open-weight alternatives economically viable for most enterprises, especially as Claude costs spiral.

— Monitor confidential TVL growth: Near's confidential Intents chain is accumulating assets across USDC, USDT, Bitcoin, ZEC, and other tokens as privacy-seeking users move funds there. This is the real privacy set for decentralized finance—watch for institutional adoption in fintech, legal, and medical sectors.

— Plan for KYC-gated AI: If major labs are nationalized or forced into KYC, decentralized alternatives using open-weight models (GLM 5.2, Llama 3, Mistral) routed through confidential inference will become the default for privacy-conscious professionals and businesses. Start exploring Near.com or similar platforms now to avoid friction later.

— Evaluate model cost arbitrage: As Anthropic's spend per customer reaches unsustainable levels, routing queries across cheaper open-weight models via confidential agents offers 90%+ capability at 90%+ cost savings—a compelling case for SME and enterprise adoption in 2025.