# Agent Platform Research Briefing — February 28, 2026
## NEW STORIES
### 1. Anthropic Pentagon Standoff — Resolved: Blacklisted, OpenAI Steps In
The Feb 27 5:01 PM ET deadline passed with Anthropic holding firm. Defense Secretary Hegseth designated Anthropic a "supply chain risk to national security," blacklisting it from working with the U.S. military or contractors. CEO Dario Amodei: "Threats do not change our position: we cannot in good conscience accede to their request." Anthropic offered to help the Pentagon transition to another provider.
Within hours, OpenAI stepped in — signing a deal with the DoD to deploy its models on the classified network. Crucially, the Pentagon *accepted OpenAI's safety red lines*, which Altman said were "similar to Anthropic's." The net outcome: Anthropic lost a $200M contract, OpenAI gained it, and the Pentagon is now on record accepting AI safety guardrails anyway. The irony is notable.
### 2. OpenClaw 2026.2.26 — Released
OpenClaw 2026.2.26 confirmed in NPM. Community references confirm the tag. GLaDOS is now four versions behind (still on 2026.2.26). Separately, Clawbot AI today launched a cloud-hosted SaaS version of OpenClaw with built-in model selection — ecosystem signal that OpenClaw's deployment model is expanding beyond self-hosted.
### 3. Rocket Lab HASTE — Mission Success
Rocket Lab successfully launched the DART AE hydrogen-fueled scramjet vehicle Feb 27. Second successful hypersonic test for the Defense Innovation Unit in three months; 7th HASTE launch overall. Australia's Hypersonix DART AE completed its maiden hypersonic flight. Clean success.
### 4. DeepSeek V4 — Still Delayed, Now Targeting March
V4 missed its February window. Now targeting March 2026. The Huawei-exclusive early optimization angle continues — US chipmakers still locked out.
---
## NO NEW DEVELOPMENTS
- Agent frameworks (LangGraph, CrewAI, AutoGen, Semantic Kernel): no notable releases
- MCP: no new CVEs or spec updates
- Voice AI: no new PersonaPlex, MiniMax, or Realtime API updates