Published today, sourced from stories within the last 24 hours.
1. South Korea commits ~$3.5B to sovereign frontier AI
South Korea plans to invest 4.7 trillion won (about $3.5B) in homegrown frontier AI starting as early as March 2027, with private capital supplementing the public outlay. The program runs two tracks — advanced model development and industrial deployment — and will open beyond current finalists LG AI Research, SK Telecom and Upstage, with roughly 29,000 GPUs allocated. (Korea Herald)
2. Google locks in 3.6 GW from Constellation, ~25% new nuclear
Google contracted 3,590 MW from Constellation Energy, with roughly a quarter from new nuclear capacity across 11 reactor upgrades in Illinois, Pennsylvania and New Jersey. The deal triggers more than $4.3B in Constellation investment, including a 20-year PPA for 890 MW of nuclear; first deliveries begin in 2028 under PJM’s ‘bring your own power’ framework for AI data centers.
3. OpenAI to test visual ads during ChatGPT image generation
OpenAI said it will begin testing a new labeled visual ad format in ChatGPT in the US later this month, displayed during image generation. With ChatGPT reaching 1.2 billion weekly users, OpenAI is integrating conversion measurement with Hightouch, Tealium and LiveRamp, with DoubleVerify and Integral Ad Science providing brand safety. Ads will stay separate from generated images and won’t influence responses.
4. Reflection AI open-sources Beam, a 501B-parameter MoE for coding
Reflection AI unveiled Beam, a 501B-parameter mixture-of-experts model with 23B active parameters, pretrained on 23.8T tokens and RL-tuned on 10.5K Nvidia GB300 GPUs. It scores 77.2 on SWE-Bench Pro v2-Hard and 80.1 on Terminal-Bench v2.1, matching Z.ai’s GLM-5.2 while using 3-4x less inference compute. Weights ship later in October under Apache 2.0.
5. Senator Schiff calls for an ‘FDA for AI’
Senator Adam Schiff said AI ‘is crying out for legislation,’ backing a dedicated regulator — ‘whether that looks like an FDA for AI or some other agency’ — plus mandatory retention of copyrighted training-data records, flagging recursive AI as a national-security concern.
6. AI execs testify under oath before NYC Council
Senior leaders from Anthropic, OpenAI, Google and Meta gave sworn testimony before the New York City Council — the first such appearance by major AI companies before a legislative body — outlining safety measures, with one witness warning that humanity could lose control of the technology. (CNBC, Qz)
7. Lisa Su asks AMD rivals to share Taiwan packaging capacity
AMD CEO Lisa Su is in Taiwan asking competitors to coordinate on advanced-packaging and substrate capacity rather than bid against each other, as AMD juggles ~14GW of AI data-center agreements including 6GW each with OpenAI and Meta. AMD has pledged $10B through 2029 to Taiwanese partners ASE, SPIL and Unimicron, and Su is extending the trip to Samsung for HBM4 memory.
8. Etched fields $40B-$50B valuation offers
AI inference-chip startup Etched is fielding bids at a $40-50B valuation, roughly double its $21B September round. The company has about $1B in secured customer orders and operates a 10MW Silicon Valley data center; Jane Street is both an investor and a customer.
9. Korea’s president orders probe after AI-driven hacks hit seven banks
South Korean President Lee Jae Myung ordered a comprehensive probe after breaches hit seven financial firms, including Shinhan Bank and KB Kookmin. Investigators found traces of ARTEX AI, a Chinese-language open-source penetration-testing framework, on infrastructure linked to the attacks.
10. Insurers brace for multimillion-dollar rogue AI claims
The Financial Times reports insurers are preparing for multimillion-dollar claims tied to rogue AI agents, with executives like Sam Altman and Dario Amodei potentially exposed under D&O liability coverage. No legal precedents yet exist for holding AI developer CEOs personally accountable for autonomous agent behavior.
11. SignSplit launches from stealth with $400M at a $1B valuation
SignSplit launched out of stealth with $400M from W Group, offering infrastructure for people to digitally sign datasets, likenesses and creative works and receive payment when AI firms use them, covering data provenance, consent terms and verification layers.
12. DayOne Data Centers files for US IPO as AI data-center revenue triples
Singapore-based DayOne filed its F-1 with the SEC seeking a Nasdaq listing. H1 2026 revenue rose to $512M from $151.5M a year earlier; the company is reportedly targeting up to $5B at a $20B valuation to fund AI data-center buildouts across Asia-Pacific and Europe.
Leave a Reply