OpenAI is rushing to raise $30 billion before Anthropic's mid-November IPO, targeting a $1.4 trillion valuation, while Meta unveiled its first Superintelligence Labs model, "Muse Wonder," intensifying the frontier race on both the financial and model fronts. Meanwhile, the three big labs confirmed they have been coordinating on cybersecurity safeguards, and in Korea, the Chinese-made AI hacking tool "ARTEX" used in a financial-sector hack was confirmed to have been used in attacks on more than 80 institutions in Japan.
💰 OpenAI Rushes to Raise $30 Billion Before Anthropic's IPO — Targeting a $1.4 Trillion Valuation
According to Reuters sources, OpenAI is rushing to raise about $30 billion ahead of rival Anthropic's mid-November listing. The valuation would exceed $1.4 trillion, with SoftBank and Abu Dhabi's MGX among those in discussions to participate. OpenAI is said to have told investors that weekly users have risen to 1.2 billion and annualized revenue to $50 billion since the launch of GPT-6, with the funds to go toward compute and data center expansion.
🧠 Meta Unveils 'Muse Wonder,' Its First Superintelligence Labs Model — Claims Frontier Parity
Meta unveiled "Muse Wonder," the first frontier model from its Superintelligence Labs, along with a larger version, "Wonder Full." The models are specialized for multi-step reasoning, coding, and agent workflows, and Meta claims their benchmark scores are on par with Anthropic's Fable 5.1 and OpenAI's GPT-6 Astra. API pricing is set at $4 per million input tokens and $20 per million output tokens.
🤝 OpenAI, Anthropic, and Google DeepMind Confirm Weeks of Coordinating Cybersecurity Safeguards
The three frontier labs confirmed they have spent weeks coordinating joint safeguards against cybersecurity-related model risks. Competition in performance, pricing, and partnerships remains fierce, but the risk of AI agents being misused for cyberattacks is seen as a "shared risk" requiring a joint response across companies. How exactly the safeguards will be implemented, and whether cooperation will continue, has not yet been disclosed.
⚡ Cloudflare's 'Clairic' — An AI That 'Extracts' Instead of Generating, Cutting Processing Time by 70%
Cloudflare has begun previewing a new AI model, "Clairic," on its Workers AI inference platform. Unlike an LLM that generates one word at a time, it extracts the needed numbers, dates, names, and other data from the input in a single pass, usually answering in under a second. Only input tokens are billed, at $0.10 per million tokens, which the company says greatly lowers the time and cost burden of processing unstructured data such as invoices and customer inquiries, where formats vary widely.
🎬 Kuaishou's Kling 4.0 — 30-Second Clips in One Go, with Stereo Audio
Kuaishou's video-generation AI, Kling, released version 4.0. It can create clips up to 30 seconds long in a single generation, accepts 10 keyframe images and prompts up to 8,000 tokens long. Two-channel stereo audio and multilingual lip-sync in more than 10 languages, including German, English, Japanese, and Korean, have also been added. A fast, low-cost "Flash" version launched alongside it, but is limited to 720p and 20 seconds.
🇰🇷 Chinese AI Hacking Tool 'ARTEX,' Which Targeted Korean Financial Firms, Also Hit 80 Sites in Japan
According to analysis by the security subsidiary of Japan's GMO Internet Group, the Chinese-made AI penetration-testing tool "ARTEX" used in attacks on Korean financial institutions was also found to have been used in cyberattacks on more than 80 organizations in Japan, including insurance, finance, and logistics firms. ARTEX is an autonomous tool that connects multiple LLMs to automate target analysis, intrusion-path planning, and security-tool execution. It was released as open source on GitHub, then switched to private by its developer after controversy. The developer now plans to relaunch it as the paid cloud service "RedAgent.ai."
🇰🇷 Samsung, SK, and LG Begin Drawing Up Next Year's Business Strategies — Focus on AI Investment Returns and Profitability
Major groups including Samsung Electronics, SK, and LG are accelerating work on their 2027 business strategies ahead of year-end management meetings. The common theme is "delivering real results from new businesses and securing profitability" after large-scale AI-focused investment. LG has begun business review meetings across all affiliates chaired by Chairman Koo Kwang-mo, with AI factories and physical AI execution plans on the agenda. SK will discuss AI semiconductor and data center investment at its CEO seminar in November, and Samsung Electronics is expected to cover HBM supply expansion and foundry profitability improvements at its global strategy meeting in December.