AI Briefing — 2026-02-28

Top Stories

Trump administration bans Anthropic from federal contracts

President Trump directed federal agencies to immediately cease using Anthropic’s AI tools with a six-month phase-out period. The action followed Anthropic’s refusal to accept Pentagon demands for “all lawful use” terms. Anthropic holds a $200 million Pentagon contract for classified military work.

Impact: 3 · Novelty: 2 · Target: 3 · Total: 8/9

Sources:


OpenAI announces agreement with U.S. Department of War

OpenAI disclosed a formal agreement with the U.S. Department of War allowing use of its AI models in classified networks. The deal includes prohibitions on domestic mass surveillance and human responsibility for use of force, with technical safeguards enforced by deployed engineers.

Impact: 3 · Novelty: 2 · Target: 3 · Total: 8/9

Sources:


Below the Fold

Anthropic’s Claude rises to No. 2 in U.S. App Store

Claude climbed to second place among free apps in Apple’s U.S. App Store, behind ChatGPT and ahead of Google Gemini. SensorTower data shows the app surged from sixth to second between Thursday and Saturday.

Impact: 1 · Novelty: 2 · Target: 3 · Total: 6/9

Sources:


Alibaba releases Qwen3.5 open-source models

Alibaba released new Qwen3.5 models (122B and 35B parameters), claiming performance comparable to Claude Sonnet 4.5.

Impact: 2 · Novelty: 2 · Target: 2 · Total: 6/9

Sources:


Tech companies announce multi-hundred-billion dollar AI infrastructure deals

Oracle signed a five-year $300 billion deal with OpenAI beginning in 2027, following a $30 billion cloud deal in June 2025. Meta plans $600 billion in U.S. infrastructure through 2028. Nvidia estimates $3-4 trillion in AI infrastructure spending by decade’s end.

Impact: 2 · Novelty: 1 · Target: 2 · Total: 5/9

Sources:


MCP server achieves 98% reduction in Claude Code context consumption

A new Model Context Protocol server reduces Claude Code’s token consumption by 98% through optimized context management.

Impact: 1 · Novelty: 2 · Target: 2 · Total: 5/9

Sources:


OpenAI opposes Anthropic supply chain risk designation

OpenAI stated via social media that it does not believe Anthropic should be designated as a supply chain risk, addressing regulatory discussions following the Trump administration’s action against Anthropic.

Impact: 2 · Novelty: 1 · Target: 2 · Total: 5/9

Sources:


Andrej Karpathy publishes minimal GPT implementation

Former OpenAI AI director Andrej Karpathy released “microgpt,” a minimal GPT implementation demonstrating core transformer concepts.

Impact: 1 · Novelty: 1 · Target: 2 · Total: 4/9

Sources:


Anthropic launches Claude memory import feature

Anthropic released a feature allowing users to import existing conversation history and context to Claude without starting fresh sessions.

Impact: 1 · Novelty: 1 · Target: 2 · Total: 4/9

Sources:


Research published on detecting LLM-generated text

Academic paper published on methods and techniques for detecting text generated by large language models, covering detection accuracy and limitations.

Impact: 1 · Novelty: 1 · Target: 2 · Total: 4/9

Sources:


Research: Minimal transformer architecture for 10-digit addition

Technical research exploring a minimal transformer architecture specifically for 10-digit addition, analyzing necessary architectural components.

Impact: 1 · Novelty: 1 · Target: 2 · Total: 4/9

Sources:


Credibility Flags

Qwen3.5 model release — Secondary source coverage only; no official announcement, paper, or repository linked from the scraped sources. Included based on high engagement scores and potential significance, but verification pending.

Run Metadata