Coverage window: articles published 2026-11-02 20:47 UTC to 2026-11-03 20:47 UTC (last 24 hours). All timestamps converted to UTC. Stories from the previous brief are excluded.
1. Anthropic’s amended S-1 adds a risk factor: it may be close to a recursive self-improvement threshold
In the revised S-1 (S-1A) filed with the SEC ahead of its IPO, Anthropic added new risk-factor language on recursive self-improvement (RSI), the point at which AI accelerates its own development. The company says it uses AI in R&D and that internal estimates suggest future (not current) models may soon meet its Responsible Scaling Policy definition of “dramatically accelerated progress attributable to AI.” It admits it has previously failed to spot a major acceleration until after the fact and could do so again, but says no current model has been designated as crossing the threshold, and that it does not expect Claude to safely design and build successive frontier models without meaningful human oversight; RSP updates are under consideration. The same filing reported roughly $17B in Q3 revenue and about $895.4B in compute and infrastructure purchase and lease commitments.
Reference: TechCrunch — 2026-11-02 22:14 UTC / Financials: AI Weekly summary (citing CNBC)
2. Nvidia reportedly tells cloud clients to sell at least 51% of services to non-Chinese companies (FT)
According to the Financial Times, Nvidia has told cloud customers running its GPUs to sell at least 51% of their services to companies outside China, effectively capping Chinese customers at 49%. The move appears to extend the tougher export-control compliance checks Nvidia applied to Asian buyers in July down to downstream cloud operators. How the ratio will be measured and enforced remains unclear, and Nvidia has not publicly confirmed the report.
Reference: Runtime Wire (citing FT and Bloomberg) — 2026-11-03 14:54 UTC
3. DeepSeek reportedly plans to pretrain V5 on 160,000 Huawei Ascend 950DT chips, a first fully domestic stack
China’s Star Market Daily, citing supply-chain sources, reports that DeepSeek plans to pretrain its next major model, V5, on Huawei Ascend chips; until now it has used Ascend mainly for inference. Unconfirmed industry tipsters on X say DeepSeek ordered about 160,000 Ascend 950DT chips (roughly $2.56B) for a 1GW data center in Inner Mongolia, with shipments ramping in Q4 and training on Atlas 950 SuperPoDs that link up to 8,192 chips. TrendForce expects V5 to arrive in 2027; neither DeepSeek nor Huawei has commented.
Reference: Wccftech — 2026-11-03 15:36 UTC / Tech City Authority (citing Star Market Daily) — 2026-11-03
4. Palantir Q3 revenue hits $2.31B (+96%) as US commercial jumps 148% and overtakes government for the first time
Palantir’s third-quarter results, released after the close on November 2, showed revenue of about $2.31B, nearly double a year earlier. Driven by demand for its AI platform, US commercial revenue rose 148% to $984M, surpassing US government revenue ($953M) for the first time, and net income reached $1.13B. Full-year revenue guidance was raised to about $8.53B (roughly 91% growth), though shares slipped about 3% in after-hours trading.
Reference: CNBC — 2026-11-02, after market close / AI Weekly summary — 2026-11-02 22:55 UTC
5. Drax, Brookfield and Radiant plan a 1GW AI data center at a North Yorkshire power station, worth up to £20B
UK power company Drax, Brookfield and Radiant will jointly develop more than 1GW of AI data center capacity at the Drax Power Station site. Total investment including chips could exceed £20B ($26.8B) if fully built; phase one, backed by an initial £2.5B, targets up to 100MW live by Q1 2028, with another 500MW planned for 2028 to 2031. The site’s existing 4GW grid connection and 2.6GW of dispatchable generation are the key draw, though no anchor tenant has been named. Drax shares rose more than 8% on the news.
Reference: CNBC — 2026-11-03 / AI Weekly summary — 2026-11-03 10:57 UTC
6. OpenAI hires former US Army Secretary Dan Driscoll as Senior Counselor for National Security Modernization
OpenAI has hired Dan Driscoll, an Iraq War veteran who resigned as Army Secretary in August after repeated clashes with the Defense Secretary, into a newly created role: Senior Counselor for National Security Modernization. He will advise leadership on AI for national security and cybersecurity, work with allied governments on “democratic AI,” and support OpenAI’s Daybreak cyber-resilience program. The hire continues OpenAI’s push into government and defense markets and policy influence.
Reference: CNBC — 2026-11-03 / AI Weekly summary — 2026-11-03 08:11 UTC
7. Apple restricts dating apps and romantic AI companion apps to users 18 and up
Apple notified developers of App Store policy changes requiring an 18+ rating for all dating and romantic-relationship apps, as well as AI companion apps offering sexual or romantic content (the Character.AI and Replika category), including apps where that is just one of many features. Apps that share a user’s real-world location with others must also disclose it, and starting November 16 developers must add new content descriptors so age ratings reflect app features. The changes are part of Apple’s broader global age-assurance and child-safety push, turning regulatory pressure on AI companions into platform policy.
Reference: TechCrunch — 2026-11-03 15:38 UTC
8. Google brings its prompt-free AI story app Dreambeans to India as US downloads fall 84%
Google expanded its experimental Dreambeans app to India, its first market outside the US, with support for English and 13 Indian languages. The app draws on connected Google services such as Calendar, Gmail, Photos and YouTube to generate illustrated stories about a user’s interests and daily life without any prompt, with some stories animated by Veo. But Sensor Tower estimates US downloads fell 84% in October to about 5,500 after a September peak, leaving demand for proactive personalized AI still unproven.
Reference: TechCrunch — 2026-11-03 03:30 UTC
Today’s Summary
- Anthropic formally flagged a self-improving-AI threshold as a risk in its IPO filing, while the Nvidia and DeepSeek reports show US and Chinese AI chip supply chains splitting further apart.
- Palantir’s results and the UK 1GW data center plan point to still-growing AI demand and infrastructure spending, while Apple’s 18+ rule and OpenAI’s national-security hire show AI moving deeper into regulation and security.
Leave a Reply