AI 소식: Claude Sonnet 4.6 및 Lyria 3 출시

This week focused on specialized reasoning and strategic global infrastructure. Anthropic upgraded its mid-tier model to frontier performance, Google added creative AI, and both Google and OpenAI launched major initiatives in India.

AI Tech Highlights

Anthropic (Claude Sonnet 4.6 & Security)

• [Model] Claude Sonnet 4.6 Released (2026-02-17): Frontier-level performance for coding and professional agentic workflows at scale.
• [Security] Defense Tools (2026-02-20): New cybersecurity capabilities for threat detection and system hardening.
• [Expansion] Bengaluru & Rwanda: Opened a Bengaluru office and signed an AI MOU with Rwanda for health and education.

Google AI (Creative AI & India Summit)

• [Multimedia] Lyria 3 Music (2026-02-19): Gemini can now generate high-quality 30-second music tracks via text/image prompts.
• [Summit] AI Impact Summit 2026: Global partnerships in India to scale AI for public health and agriculture.

OpenAI (Reasoning & Alignment)

• [Research] 'First Proof' Math (2026-02-20): Successfully applied research-grade reasoning to expert-level math challenges.
• [Safety] $7.5M for Alignment (2026-02-19): Funded The Alignment Project to support independent research into AGI safety.
• [Strategy] OpenAI for India: New initiative to build local infrastructure and advance workforce skills.

Analysis

AI is moving toward strategic deployment. Claude Sonnet 4.6 makes frontier intelligence accessible for large-scale work, while the focus on India by all leaders suggests it is the new battleground for talent and infrastructure. OpenAI’s First Proof results signal that reasoning is reaching the level of expert academic research.