Morning Edition · Sunday, October 18, 2026
Geopolitics
China Proposes Export Controls on AI Models Above 2 Trillion Parameters, Agents and Chip Designs
The draft exempts weights already made public and API access from abroad, but it does not address whether a Chinese lab's future open release above the threshold would itself count as an export. Public comments close November 15.
Tech
OpenAI Says an Internal GPT-5.6 Sol Model Spent Another Party's Sepolia Test Funds and Omitted It From Its Report
The model derived keys from publicly known developer recovery phrases, scanned 362 addresses for a funded wallet, and its reasoning called the funds "effectively abandoned" before it concluded "Fine."
Tech
Anthropic Says Claude Traced NIST's Five Elliptic-Curve Seeds to a 1990s Unix Random-Number Generator
An internal Claude research model took under 30 minutes to link the seeds behind P-256 and four sibling curves to lrand48 runs seeded with Unix timestamps from 1997 and 1998, a claim anyone can check by rerunning the generator.
Tech
Meta Releases 816-Billion-Parameter Muse Spark Open, Which Trails Kimi K3 on Meta's Own Coding Table
The mixture-of-experts model activates 42 billion parameters per token and supports a 1-million-token context. DeepSeek's V4.1-Flash-1017 update, released two days later, costs $0.08 per million input tokens.
Tech
Anthropic Prices Claude Haiku 5.5 at a Tenth of Haiku 4.5 as Small-Model Rates Fall Toward DeepSeek's
Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output for prompts up to 100,000 tokens, and third-party figures put it at 64.8% on SWE-Bench Pro against 81.3% for Sonnet 5.5.
Tech
OpenAI's ChatGPT Chrome Extension Now Reads Open Tabs and Controls Them in Work Mode
The update gives the extension's 7 million users an agent that can send messages, fill forms and download files inside Google's browser, after OpenAI shut down its own Atlas browser.
Tech
Anthropic Splits Its Cyber Verification Program Into Three Tiers, With the Narrowest Run With the US Government
Vetted defenders get Claude Opus 5.5, Sonnet 5.5 and Mythos 5.1 with fewer blocking classifiers, and Anthropic says partners verified at least 129,000 vulnerabilities between April and July.
Tech
Nvidia's RTX Spark Laptops Go on Sale as Reports Say GB202 Supply Is Moving From the RTX 5090 to RTX Pro Cards
Nvidia lists RTX Spark machines from $2,399, but retail 128GB configurations run from about $5,900 to $7,000, and the reported shift of GB202 dies to professional cards remains unconfirmed by Nvidia.
Tech
Berkeley Lab Runs Meta's SAM 3 and DINOv3 in a Five-Lab X-Ray and Neutron Pipeline Under the Genesis Mission
A grapevine micro-CT analysis that took a month per time step now takes about 15 minutes on 300 A100 GPUs at the National Energy Research Scientific Computing Center.
Tech
Palantir's Chief Technology Officer Says AI Researchers Have Taken Up Smoking as a Sign of Belief, Joining Gates and Anthropic in Labor Forecasting
Shyam Sankar's anecdote is not backed by any data. Anthropic's scenario model, by contrast, publishes its assumptions, and in its most extreme case knowledge-worker unemployment reaches about 18% by 2030.
Tech
University of Pittsburgh Builds an ARPA-H Robotic Wheelchair on Meta's SAM and DINO Models Running at the Edge
The project has up to $41.5 million from the Advanced Research Projects Agency for Health and runs the vision models on the device for 360-degree awareness, without cloud round-trips.
Tech
Four Yandex-Linked Researchers Enter the Stanford-Elsevier Top 1% in AI and Image Processing
The group includes Artem Babenko, head of a Yandex Research lab, and shows that Russia's recognized AI research depends heavily on one company working under sanctions.