The KOSPI plunged 5.28% on concern that the Democrats' midterm "blue wave" could bring tighter AI data center regulation, and Anthropic agreed to rent capacity in SpaceX's massive data center. OpenAI released image generation 2.6, and Naver was hit with a copyright compensation claim from broadcasters over AI training. We also round up Tencent's Hy4, the unified AI scraping opt-out, and Google DeepMind's chemistry agent.
🇰🇷 KOSPI Plunges 5.28% to 5,383.80… Samsung Electronics -7.57%, SK Hynix -9.46%
As concern grew over tighter AI data center regulation following the Democrats' sweeping midterm victories, the KOSPI closed at 5,383.80, down 300.00 points (5.28%) from the previous session. Foreign investors net sold more than ₩3.4 trillion, and Samsung Electronics (-7.57%, ₩201,500) and SK Hynix (-9.46%, ₩1,264,000) plunged together. Sell-side circuit breakers were triggered on both the KOSPI and KOSDAQ markets, and the KOSDAQ also closed down 5.11% at 678.41.
🖼️ OpenAI Releases Image Generation 2.6… GIF Support and a Fast API Model
OpenAI released ChatGPT image generation 2.6 on web, desktop, and mobile. New features include native animated GIF generation and a reusable reference-image pinning function, and for the API it added the fast Flare model and Sunburst, which can handle SVG. Weekly image generation reportedly now exceeds 4 billion images.
🚀 Anthropic Signs Deal to Use SpaceX's Massive Data Center… Secures 300MW at Colossus 1
Anthropic has signed a data center usage agreement with Elon Musk's SpaceX. The Colossus 1 facility in Memphis, Tennessee, has more than 220,000 Nvidia chips, and Anthropic will secure 300 megawatts of new compute capacity within a month. Musk said he decided to lease the capacity after spending time with Anthropic's leadership last week, and the two companies also unveiled a new Claude feature, "Dreaming," at a developer event.
📰 Naver Faces Copyright Compensation Claim Over AI Training… Three Broadcasters Demand ₩15 Billion
Three Korean broadcasters have claimed ₩15 billion (about $11.16 million) in compensation from Naver and Naver Cloud. The broadcasters allege that Naver used their news content without authorization years ago to develop AI models. The case signals that copyright disputes over data used for AI training are heating up in Korea, and the eventual precedent is being watched closely.
🐉 Tencent Unveils New Model Hy4… Visual Reasoning, GUI Automation, and Long-Horizon Agents
Tencent unveiled a new model, Hy4, on November 4. Its core features target image understanding, computer screen manipulation (GUI automation), and long-horizon agent workflows. However, the promised model weights and safety report have not yet been released, so verification of its actual performance and safety is still pending.
🤖 OpenAI, Google, Microsoft, and Anthropic Back Unified AI Scraping Opt-Out Signal
OpenAI, Google, Microsoft, and Anthropic have agreed to support a unified AI scraping opt-out signal, the "policy token," placed in robots.txt. The approach unifies rules that previously differed from crawler to crawler into a single token; Anthropic has already implemented it, OpenAI plans to do so within months, and Google and Microsoft have announced 2027 rollouts. Standardization is being pursued through the IETF AIPREF working group.
🔬 Google DeepMind's Chemistry Agent 'Antigravity'… Proposes Drug Candidate Molecule in About an Hour
Google DeepMind said the chemistry skill in Antigravity can analyze compound data and propose a macrocycle candidate in about an hour. The designed molecules still need actual synthesis and laboratory validation, but the case shows AI agents can greatly accelerate the early stages of drug development. Meanwhile, Google's Gemini 4 Carbon edged out Claude Opus 5.5 by one point on the Artificial Analysis Intelligence Index.