U.S. Bans DeepSeek; Arm's AGI CPU Outperforms AMD

Key Takeaways

• U.S. DHS orders all federal agencies to remove DeepSeek within 60 days over data security and national security concerns
• China's Foreign Ministry and Commerce Ministry call the DeepSeek ban discriminatory and politicize technology claims
• Arm claims its AGI CPU delivers double the performance per rack versus AMD's EPYC 9755 across seven workloads
• Gigamon launches GigaVUE-GSA110 for AI network monitoring at up to 800 Gbps optimized packet delivery
• Annexa's Erstan platform cuts procurement order processing from hours to about five minutes for NetSuite users
• UCLA supercomputer analyzes 100,000 breast cancer patient records to train more trustworthy diagnostic AI, published in Nature Medicine
• Harvard's Christopher Stubbs joins Anthropic's Long-Term Benefit Trust; Douglas Finkbeiner named his successor
• ECLAC and Chile's CENIA release ILIA 2026 report urging Latin America to invest more in AI research and address job displacement
• DeepSeek faces global scrutiny after its R1 model sparked a Nvidia sell-off in January 2025
• Corelight's Vince Stoffer warns security teams must maintain guardrails and human approval over AI-driven response actions in SOCs

China pushes back against U.S. ban on DeepSeek AI

The U.S. Department of Homeland Security ordered agencies to remove DeepSeek products from government IT systems. The order claims DeepSeek models create security risks and that Chinese authorities could access U.S. data. China's Foreign Ministry spokesperson Mao Ning said the U.S. is overstretching national security concerns and will defend Chinese companies. China's Ministry of Commerce also opposed the ban and said the U.S. is trying to suppress China. The dispute comes as both nations seek some cooperation on AI governance.

US bans DeepSeek from all federal agencies over security risks

The US Secretary of Homeland Security issued an order banning DeepSeek products and services from all federal agencies. Government agencies must remove DeepSeek from their IT systems within 60 days. The ban cites concerns about DeepSeek servers in China, data collection on users, and possible security vulnerabilities. Agencies can request exemptions by showing compelling justification. The ban does not apply to private consumers or the Department of Defense. DeepSeek recently open-sourced its DeepInfer framework for AI inference.

China criticizes U.S. ban on DeepSeek AI in government systems

The U.S. Department of Homeland Security ordered civilian government agencies to remove DeepSeek products and any products built on its open-source models. The order also bans future purchases of DeepSeek services. U.S. officials worry that Chinese law gives the government access to company infrastructure and sensitive data. China's Ministry of Commerce called the ban discriminatory and said China has laws protecting cybersecurity and data privacy. China urged the United States to withdraw the restrictive measures.

US orders federal agencies to remove DeepSeek over national security

The U.S. Department of Homeland Security ordered federal civilian agencies to stop buying from DeepSeek and remove its products from government systems. The order also covers third-party products built using DeepSeek technology. DHS warned that Chinese authorities could access sensitive U.S. government data under Chinese law. China's Foreign Ministry spokesperson Mao Ning opposed broadening national security claims and politicizing technology. DeepSeek faces global scrutiny after its R1 model sparked a Nvidia sell-off in January 2025. A U.S. report accused DeepSeek and other Chinese AI firms of large-scale model distillation.

Gigamon launches new appliance for AI network monitoring

Gigamon released the GigaVUE-GSA110, a new GigaSMART Network Appliance for monitoring AI-enabled networks. It supports AMI across up to 400 Gbps of traffic and can deliver optimized packet data for up to 800 Gbps. The device fits in a compact 1RU size and uses a disaggregated architecture for flexible processing. It helps organizations track sanctioned and unsanctioned AI use across hybrid cloud environments. Gigamon says the tool gives security teams trusted network context for AI operations.

Annexa expands AI tools for NetSuite customers at SuiteWorld 2026

Annexa is building AI agents for Oracle NetSuite users in Australia and New Zealand. Its Erstan platform lets AI agents handle tasks in procurement, inventory and sales. A procurement assistant for retailer Boody cuts order processing from three to six hours down to about five minutes. SuiteWorld 2026 in Las Vegas will showcase new AI features inside NetSuite. Director Matthew Owens says AI is shifting from answering questions to completing real workflows.

AI in security operations centers needs guardrails against shadow AI

Vince Stoffer from Corelight explains how AI helps security teams work faster. LLMs assist with writing SIEM rules and analyzing network data. Agentic triage can reduce false positives and help junior analysts dig deeper. However, a human must still approve any blocking or response actions. Attackers also use AI for coordinated attacks and zero-day exploits, so teams must watch for shadow AI and loose permissions.

Career advice for students entering the AI era

A speaker at National Taiwan University told students that the career ladder has changed but not disappeared. AI now handles much of the work that once filled entry level jobs, so young workers will direct tasks sooner. The speaker advised students to build their personal skills and health rather than chase job titles. Asking good questions and owning the answers matters more than ever. The playbook for careers in the AI age is still unwritten.

AI and social media reshape democracy in the election era

An opinion piece reflects on how AI and social media have changed politics across generations. The 2024 election combined social media platforms like Facebook and Twitter with AI tools that helped candidates target voters. The author describes this as a semiquincentennial of apathy driven by algorithms. The article asks whether these tools will spread misinformation or encourage more engaged citizens. Ultimately the author says the choice about how to use AI rests with the public.

Supercomputer aids development of reliable AI for breast cancer diagnosis

UCLA researchers built a supercomputer to make AI more trustworthy for diagnosing breast cancer. The machine analyzed data from 100,000 breast cancer patients and found features that predict the disease. These features were used to train an AI model that can accurately diagnose patients. The study was published in Nature Medicine. The approach could improve the reliability of AI tools in healthcare.

Latin America and the Caribbean urged to take ownership of artificial intelligence

ECLAC and Chile's National Centre for Artificial Intelligence (CENIA) released the fourth edition of the Latin American Index of Artificial Intelligence (ILIA 2026). The report says the region has made progress in AI adoption with many countries creating national AI strategies. However, it calls for more investment in AI research and better policies and regulations. The report also warns about social and economic impacts like job displacement and the need for workers to learn new skills.

Former Harvard AI Adviser Joins Anthropic Oversight Trust

Harvard Physics and Astronomy professor Christopher W. Stubbs joined Anthropic's Long-Term Benefit Trust. This oversight body can appoint most of Anthropic's board and advises on AI risks and societal impacts. Stubbs stepped down as the Faculty of Arts and Sciences' senior adviser on AI less than three weeks before the announcement. He remains an active Harvard faculty member and continues serving on the FAS committee on generative AI. Astronomy and Physics professor Douglas P. Finkbeiner was named as his successor.

Arm claims AGI CPU delivers double the performance per rack of AMD EPYC 9755

Arm developed a new CPU evaluation methodology for measuring agentic AI performance across seven workloads. The methodology was developed with support from Fujitsu, Meta, and OpenAI. Arm tested its AGI CPU against AMD's EPYC 9755 and claimed double the performance per rack. The tests covered tooling, agentic workflows, reasoning and planning, and sandbox-creation overhead. Arm used a 128-core and a 136-core AGI CPU configuration while AMD used one socket of the 128-core EPYC 9755.

Sources

• China defends DeepSeek after U.S. ban of its AI from government systems
• US bans DeepSeek across federal agencies, putting AI supply chains under scrutiny
• China Criticizes U.S. Ban on DeepSeek A.I. in Government Systems
• U.S. Orders Federal Agencies to Remove DeepSeek AI Products Over National Security Concerns
• Gigamon Extends the Deep Observability Pipeline for AI-Enabled Network and Security Operations with New GigaSMART Network Appliance
• Annexa targets new AI opportunities at SuiteWorld 2026
• Bringing AI into the SOC means watching for shadow AI and loose permissions
• The Ladder Isn't Gone: My Advice to Students Starting Their Careers in the AI Era
• A Semiquincentennial of Apathy … and AI
• Supercomputer helps design more reliable AI for breast cancer diagnosis
• Latin America and the Caribbean urged to take ownership of artificial intelligence
• Former Harvard AI Adviser Joins Anthropic Oversight Trust Weeks After Stepping Down
• Arm claims AGI CPU delivers double the performance per rack of AMD EPYC 9755 in independent analysis — new CPU evaluation methodology developed by Arm evaluates agentic AI performance across seven workloads