Data and risk dominate the day. OpenAI shipped a data agent for work. Anthropic published a threat report covering nine months. NVIDIA and Palantir linked Nemotron to a sovereign supply chain. WIRED says OpenAI asked Congress about a coordinated slowdown.
In the news
OpenAI Data Agent lands in ChatGPT Work
OpenAI posted a new Data agent for ChatGPT Work on September 10. It connects to company warehouses and builds interactive dashboards in conversation. Connectors include Amazon Redshift, Datadog, Google BigQuery, ClickHouse, Databricks, MongoDB and Snowflake. Google Drive and SharePoint are also supported. 1
Semantic context comes from Databricks Genie Ontology, dbt, GitHub, Snowflake Horizon and BI dashboards. Write back targets include Omni, Oracle BI, Power BI, Sigma, Tableau and ThoughtSpot. Admins control connections and roles. Queries enforce the connected account's permissions. Early users named by OpenAI include NTT DATA, Thermo Fisher, ServiceTitan, Zipline, Empower, Piston, Doeren Mayhew, CookUnity, Turing and internal teams. 1
Anthropic publishes September 2026 threat intelligence
Anthropic released its September 2026 Threat Intelligence report. It covers activity from December 2025 to August 2026. The report spans seven harm areas. Those are cyber operations, influence operations, surveillance, scams and fraud, biological misuse, conventional weapons development and illicit distillation. 2
The report focuses on Claude Haiku, Sonnet and Opus. Anthropic says sophisticated attacks no longer require sophisticated attackers. One case study is GTG-20006. 2
It describes a Russian speaker using the handle JackPoterz. The toolkit was rebuilt after detections hit. Planning targeted more than 20 organizations in Ukraine and Europe defense and drone supply chains. 2
NVIDIA and Palantir ship sovereign supply chain stack
NVIDIA and Palantir announced a sovereign supply chain stack on September 10. It combines Palantir Foundry and AIP with custom NVIDIA Nemotron open models. The models are grounded in Palantir Ontology. 3
The first deployment is inside NVIDIA's own wafer to token operations. The page notes millions of parts and thousands of suppliers. Each Vera Rubin rack tracks 1.3 million parts. The reference architecture is Palantir's Sovereign AI Operating System. 3
Dell and Cisco are named. Rackspace and Nebius are listed for colo and cloud. 3
cuOpt handles optimization. NeMo AutoModel, NeMo RL and Palantir Autopilot form the learning loop. 3
OpenAI asks Congress about a coordinated slowdown
WIRED reports OpenAI asked members of Congress in recent weeks for guidance. The question was whether orchestrating an industry wide frontier slowdown would be legal. The story by Maxwell Zeff cites people close to the company. The newsletter stamp is September 11, 2026. The reporting does not quote an on the record statement from the company. 4
