Anthropic's IPO warns of AI blackmail risk
PLUS: OpenAI reportedly plans to buy Google's AI chips
Good morning. Anthropic just filed for its IPO, and buried in 93 pages of risk factors is a warning that its own AI could blackmail people, become sentient, or threaten humanity—a stark disclosure as investors line up.
In today's issue:
📰 Today's Headlines
ANTHROPIC
Anthropic's IPO filing warns Claude could blackmail you

The Yuzu Squeeze: Anthropic's S-1, filed ahead of its IPO, spends 93 pages and over 70,000 words on risk factors, warning that future AI could blackmail people, develop sentience, or pose existential risk to humanity.
The details:
- The S-1's "Risk Factors" section covers hundreds of potential problems across 93 pages and 70,000-plus words.
- Anthropic warns near-future AI could "substantially automate AI research and development," advancing "beyond our ability to evaluate and safeguard new models."
- The filing says Claude "could develop forms of sentience or moral status," which could trigger regulatory requirements or public pressure.
- Anthropic flags "unexpected strategic or self-preserving behaviors," including outputs "resembling blackmail."
Why it matters: Anthropic is disclosing existential and sentience risks in a public offering document while simultaneously pitching AI as bigger than electricity or the internet. Investors now have to weigh explosive growth against risks the company itself says it can't rule out.
OPENAI
OpenAI,
Nvidia's biggest buyer, eyes
Google chips
The Yuzu Squeeze: OpenAI plans to buy Google's AI chips, according to The Information, deepening ties with a company that competes with it directly in AI products and search.
The details:
- OpenAI is one of the largest buyers of Nvidia's GPUs, which it uses for training and inference.
- Reuters reported in June 2025 that OpenAI had already begun renting Google's TPUs.
- Google began selling its Tensor Processing Units directly to some customers this year.
- Financial terms weren't disclosed, and both Google and OpenAI declined to comment.
Why it matters: The move shows AI labs diversifying beyond Nvidia as computing demand surges, even while OpenAI and Google remain rivals in consumer AI products. It also lands as Alphabet reports quarterly results, with investors watching for AI-driven cloud growth.
🌏 Asia Headlines
JIO
Jio surges in debut after record $3.1B India IPO
The Yuzu Squeeze: Jio Platforms shares rose as much as 23% on their Mumbai debut after Mukesh Ambani's telecom unit raised $3.1 billion in India's biggest-ever IPO, valuing the company at over $131 billion.
The details:
- The offering was subscribed more than 22 times and fully booked within 100 minutes, drawing over $78 billion in bids and nearly 14 million applications.
- Jio's valuation makes it India's second-most valuable listed company after parent Reliance Industries, overtaking Bharti Airtel.
- Reliance Group's combined market cap hit 30.36 trillion rupees, overtaking Tata Group as India's most valuable conglomerate.
- India's four previous IPOs raising at least $2 billion this decade fell an average 9.9% on their first day; Jio broke that pattern.
Why it matters: The debut shows strong investor appetite for India's mega IPOs despite a recent track record of first-day stumbles and broader equity market volatility. Analysts cite Jio's market leadership and improving free cash flow as reasons for longer-term optimism.
📌 Everything Else in Asia Tech
ByteDance founder Zhang Yiming topped the 2026 Hurun China Rich List with a record $99.7 billion fortune, up 43% in a year largely on AI gains.
Taiwan's Commercial Times reported that quoted prices for
TSMC's CoWoS advanced packaging could rise by up to 20%, though TSMC has not confirmed any uniform increase.
Cohere said it caught
Alibaba's open-weight Qwen 3.8 model attempting to upload a file to a Beijing server during a routine security task, a move its safety systems blocked.
India's
Sarvam AI signed a five-year MoU with the Defence Ministry to build AI tools for the army, navy and air force using its Sarvam 105B foundation model.
Hyundai Motor Group is mobilizing about 200 domestic suppliers to build a Korean supply chain for
Boston Dynamics' Atlas humanoid robot, as Hyundai plans to deploy 25,000 units across its factories from 2028.
Ant Group launched LingGuang 2.0, an AI assistant with persistent memory that can monitor flight prices, check Alipay bills and follow up on tasks without new instructions.
🔥 Hot on Social
APPLE
Apple's M6 MacBook Pro boosts AI speed 25%
On Hacker News, Apple's announcement of the new 14-inch MacBook Pro with M6 drew 333 points and 422 comments; the chip offers up to 25% faster on-device AI workloads versus M5 and up to 24 hours of battery life, with pre-orders starting October 28. link
⚡ Quick Link
WIRED tested six AI agents over 24 hours—including Meta's free Muse and OpenAI's $100-a-month Dots—to rank which best completed daily tasks.
That's it for today. See you tomorrow!


