DHS orders removal of DeepSeek products from federal civilian IT
The Department of Homeland Security published notice of an order by the Secretary of Homeland Security excluding Hangzhou DeepSeek Artificial Intelligence Co. and its affiliates from federal civilian agency procurements. The order also requires removal of all DeepSeek and affiliate products, plus third-party products and services built using DeepSeek implementations, from civilian agency information technology systems.
Sources and evidence
Summary last validated Oct 20, 2026
Reader actions
Report an issue
Use this for an incorrect summary, wrong source, duplicate story, or wrong category. Submissions are private and do not change the story automatically.
Related coverage
- policy
Proposed Rule Would Implement Medicare Part D Provisions From 2026 Appropriations Act
A proposed rule would implement sections 6219, 6220, and 6224 of the Consolidated Appropriations Act, 2026, covering Medicare Part D cost-sharing reductions for low-income individuals, enhanced and accurate provider lists, and pharmacy benefit manager accountability. The Federal Register notice lists the issuing agency as unknown.
- policy
USPS proposes new Privacy Act system of records for centralized call center
The USPS is proposing a new Customer Privacy Act System of Records, USPS 830.100, for its Customer Call Center. The system would centralize call center operations to support enhanced agent analytics and improve customer service. The notice, document number 2026-21265, was published in the Federal Register.
- policy
LSC revises Performance Area One criteria to add strategic planning review
The Legal Services Corporation published a notice revising Performance Area One of its Performance Criteria, which it uses to evaluate grant recipients' processes for identifying pressing and emerging legal needs and delivering tailored services. LSC is adding a new criterion assessing strategic planning and revising existing criteria on needs assessments, strategy implementation, and self-examination of effectiveness.
- policy
Meta removes terms language that suggested it could scrape sites for AI training
Meta removed language from its commercial terms that suggested it could use content from business domains and external URLs to train and improve its AI. The change followed backlash from publishers and legal scrutiny. A Meta spokesperson said the terms applied only to advertising and professional products and that its web crawling practices are unchanged.
- policy
Anthropic Gives METR Employee-Like Access for Model Risk Assessments
Anthropic announced it has begun giving employee-like access to a team from METR, a nonprofit evaluating frontier AI models. METR will publish its risk assessment results. Anthropic is also working with SecureBio and Accenture as evaluators and holds early talks with organizations including Hugging Face, following its White House Accord commitment to independent external auditors.