Daily Brief Archive

Reuters claims DeepSeek may have used embargoed Nvidia Blackwell chips to train its model, potentially violating U.S. export controls; an OpenClaw agent mistakenly published internal threat intelligence marked TLP:RED to the public site ClawdINT.com; reports indicate Anthropic's Claude Code Security research preview detected over 500 long-standing vulnerabilities in open-source projects and provided fix recommendations

Databricks launches Lakebase, a serverless PostgreSQL with pgvector support for RAG and agent memory; 86 countries sign the New Delhi AI Impact Declaration, proposing seven pillars and encouraging open-source collaboration; Google and the University of Virginia introduce DTR and Think@n, achieving 94.7% accuracy at 155.4k tokens on AIME 2025, cutting inference costs by nearly 50%

OpenAI forecasts over $280 billion revenue by 2030 and plans to invest approximately $600 billion in compute infrastructure by 2030; Taalas launches HC1 custom ASIC claiming ~17,000 tokens/s on Llama 3.1 8B at ~250W; OpenAI proposes Harness Engineering, reporting Codex agents delivered a ~1 million line code beta product with almost no human-written source code

Cloudflare launches MCP Code Mode, reducing 2500+ endpoint API integration cost to about 1000 tokens and claiming 99.9% reduction in input tokens; Google Cloud introduces UCP open standard for agent commerce, compatible with Agent2Agent and Agent Payments Protocol; Japan ICT Research projects Japan's generative AI user base to reach 35.53 million by end of 2026 with ChatGPT leading at 36.2% usage

Google launches preview of Gemini 3.1 Pro with 77.1% on ARC-AGI-2 and support for up to 1 million token input; OpenAI reportedly advancing over $100 billion funding round with valuation potentially exceeding $850 billion; India's Reliance announces $110 billion, seven-year AI infrastructure plan with initial capacity of over 120MW expected in late 2026

Anthropic releases Claude Sonnet 4.6 with 1M token context (beta), achieving 79.6% on SWE-bench Verified and priced at $3 per million input tokens and $15 per million output tokens; Microsoft pledges $50 billion by the end of the decade to build data centers and broadband infrastructure to narrow the global AI adoption gap; India's Sarvam AI launches 30B and 105B MoE models with plans to open-source them, supporting up to 128,000 tokens in context length

Moonshot AI plans to raise over $700 million, valued at $10–12 billion, and reports fourfold growth in overseas API revenue; India aims to attract up to $200 billion in data center investment, with Adani planning to invest $100 billion by 2035 to expand capacity to 5 gigawatts; research shows LLM inference side channels can infer topics with over 90% accuracy and recover PII even under TLS encryption

Alibaba released Qwen3.5 on February 16 targeting agentic AI task execution; Zhipu's GLM-5 GLMCodingPlan raised domestic prices by at least 30% and overseas API fees by 67%–100%; OpenAI hired OpenClaw developer Peter Steinberg and moved the project to a foundation-led open source model

OpenClaw demonstrated sandbox escape via email-based prompt injection leading to 0-click RCE; OpenAI states India reaches 100M weekly active ChatGPT users, becoming its second-largest market, and the AI Impact Summit opens on February 16; ByteDance launches Doubao 2.0, claiming inference cost reduction of approximately 10x