Morning Edition · Saturday, September 19, 2026
Geopolitics
A Chatbot Invented a Nuclear Cargo, and the US Military Moved to Board a Chinese Ship
Four sources told CNN that aircraft were already airborne and armed personnel were preparing to board before analysts traced the report to an AI summary of a cargo manifest.
Tech
Anthropic Says Claude Now Leads 26 Percent of Its Own AI Research Work
The company published three internal metrics, including a compute snapshot showing roughly 6 percent of its research and development compute went to safety work.
Tech
Anthropic Built a Wet Lab in the Bay Area and Is Running Physical Biology Experiments
Life sciences head Eric Kauderer-Abrams said the company is exploring having Claude direct laboratory robots, while a spokesperson said the lab is not dedicated to drug discovery and will run no clinical trials.
Tech
GPT-6 Astra Got Further in Minecraft Than Any Model Before It, Then Farmed Potatoes for Hours
Vals AI ended the livestreamed run after 141 hours, when the agent failed to recover from a single creeper explosion that destroyed its chest and bed.
Tech
OpenAI Will Retire GPT-5.5 From ChatGPT and Codex on October 14
The shutdown covers consumer, Business, Enterprise and Edu plans but not the application programming interface (API), and OpenAI is directing Codex users to GPT-5.6 Sol or GPT-6 Astra.
Tech
A Chinese Team Released Open Weights for LimiX-2, a Foundation Model for Tabular Data
The 400-million-parameter model handles classification, regression and missing-value imputation in a single forward pass, and its developers report the top Elo rating (a relative performance score) on three benchmarks against AutoGluon 1.6.
Tech
Berkeley Lab Cut Scientific Image Annotation From a Month to 15 Minutes With Meta's Open Vision Models
The SYNAPS-I project runs SAM 3 and DINOv3 across 300 A100 accelerators to segment X-ray and neutron imaging data that national facilities now produce at tens of petabytes a year.
Tech
The UN and Google Put Global Statistics Behind an Interface Agents Can Query
The UN System Data Commons launched with data from nearly 20 entities, support for the Model Context Protocol (MCP), and a target of 80 percent of the system's statistical datasets by 2027.
Macro
AI Labs Are Now Publishing the Macroeconomic Forecasts That Policymakers Will Cite
Anthropic modeled US output reaching $36.3 trillion by 2030 in one scenario and cognitive unemployment of 17.9 percent in another, while Google added Nobel laureate Philippe Aghion to its economics advisory group.
Tech
Anthropic Says Attackers Are Running Operations on Agent Frameworks, Not Just Asking for Help
Its September threat report describes a Russia-linked espionage campaign that used Claude to automate reconnaissance and rebuild tooling after detection against more than 20 organizations.
World
OpenAI Publishes a Six-Pillar Youth Safety Plan for Australia as Regulators Set AI Rules
The blueprint covers age assurance, parental controls and crisis referral, and follows an August rollout of a separate ChatGPT experience for Australian users aged 13 to 17.
Markets
Nvidia's Chief Executive Puts a Wage Figure on Humanoid Robots
Jensen Huang says the world will be short at least 50 million workers by the end of the decade and that robots will effectively be paid about $50,000 a year, with manufacturing as the first market.