# Agent Platform Research Briefing — March 10, 2026
Generated: 2026-03-10 06:00 AM PT
---
## NEW STORIES
### 1. Anthropic Launches Claude Code Review — Multi-Agent PR Analysis in Research Preview
**Sources:** Dataconomy (Mar 10), India Today (Mar 10), Dev.to (Mar 10), claude.com/blog/code-review
Anthropic launched **Code Review**, a new feature within Claude Code, on March 10, 2026. It is currently in **research preview** for Claude for Teams and Claude for Enterprise customers.
- Automatically analyzes GitHub pull requests and posts comments on potential issues
- Focuses on **logic errors** rather than style issues, using a color-coded severity system: red (highest priority), yellow (potential issues), purple (preexisting code problems)
- Uses a **multi-agent architecture**: multiple agents examine the codebase in parallel, then a final agent aggregates, deduplicates, and ranks findings
- Anthropic's enterprise growth has been extraordinary: daily signups have **quadrupled since January 2026**, with Claude Code run-rate revenue surpassing **$2.5 billion**
- Anthropic claims engineers marked **less than 1%** of findings as incorrect — unusually low for automated review tooling
- Pricing is token-based: estimated average cost of **$15–$25 per review** depending on complexity
- Deeper security analysis is available via the separately-announced Claude Code Security preview
**Context:** The launch explicitly addresses the bottleneck created by AI-generated code volumes overwhelming human reviewers — a problem that scales with enterprise AI coding adoption. Customers like Uber, Salesforce, and Accenture are cited as targets. This is a direct step toward Anthropic's vision of fully autonomous software engineering pipelines.
---
### 2. OpenAI Acquires Promptfoo — Agentic Security Testing Comes to OpenAI Frontier
**Sources:** TechCrunch (Mar 9), CNBC (Mar 9), SiliconAngle (Mar 9), Help Net Security (Mar 9)
OpenAI announced plans to **acquire Promptfoo**, an open-source AI security testing platform, on March 9, 2026. Financial terms were not disclosed. Promptfoo had raised **$23.6 million** total ($18.4M from a July round).
- Developer-focused security testing for LLM-based systems
- Detects: prompt injections, jailbreaks, data leaks, tool misuse, out-of-policy agent behaviors
- Provides an open-source interface and library used widely in enterprise dev pipelines
- Promptfoo's technology will be integrated into **OpenAI Frontier**, the enterprise platform for AI agents (formerly "AI coworkers")
- Automated security tests will run during development, not post-deployment
- Monitoring, documentation, and compliance reporting will be added as core features
- The Promptfoo team joins OpenAI
- Signals that major AI labs are now treating **agentic security testing as a core product requirement**, not a third-party concern
- Promptfoo is widely used (open-source with enterprise tooling) — pulling it inside OpenAI could fragment the independent testing ecosystem
- The acquisition comes amid high-profile agentic security concerns: MCP deployment vulnerabilities, Claude Code RCE disclosures, and the ClawHavoc supply-chain attack all surfaced in the past three weeks
---
### 3. Anthropic Files Two Lawsuits Against the Department of Defense
**Sources:** TechCrunch (Mar 9), NPR (Mar 9), NYT (Mar 9), Wired (Mar 9), Bloomberg (Mar 9)
Anthropic filed **two federal lawsuits** against the Department of Defense on Monday, March 9, challenging the Pentagon's "supply chain risk" designation under 10 U.S.C. § 3252.
- The lawsuits are now **filed** (as of March 9) — previously this was announced as Amodei's intent to challenge
- The complaint calls DoD's actions "unprecedented and unlawful"
- Multiple Anthropic customers have reportedly said they are **pursuing alternatives** due to the designation — creating real commercial pressure
- OpenAI and Google employees have publicly expressed support for Anthropic's position
- Anthropic faces losing **hundreds of millions of dollars** in annual government revenue plus downstream B2B revenue from federal contractors who use Claude
**Context:** The designation was made March 5. Anthropic launched Claude Code Review today — Anthropic itself says the enterprise business is "critical" in the context of this dispute. Both stories are intertwined: enterprise revenue growth is the financial cushion for the legal fight.
---
## CONTINUING STORIES (Significant Updates)
### Firefly Alpha "Stairway to Seven" — Seventh Scrub, March 10 Window Tonight
- March 9 attempt scrubbed due to an **"out of range" sensor reading** (NASASpaceFlight.com, Mar 9)
- New window: **5:50–7:50 PM PDT tonight, March 10** (00:50–02:50 UTC March 11)
- This is the seventh launch attempt for FLTA007; mission is a return-to-flight after the failed FLTA006 "Message In A Booster" in April 2025
- Still awaiting first successful Alpha flight since the mission failure
### OpenClaw — 2026.3.8 Remains Latest, No New Release
- GitHub releases page confirms v2026.3.8 (released March 8-9) is still the current stable release as of March 10
- No new OpenClaw release today
---
## SKIPPED (Not New)
- OpenClaw security history (ClawJacked, Endor CVEs) — resolved/covered
- DeepSeek V4 — still unreleased (do not resurface until confirmed)
- Grok racist posts investigation — no root cause disclosed yet
- Starship Flight 12 — static fire window open (March 8-10), no confirmed result as of 6 AM PT
---
*End of briefing. 3 new stories, 1 update.*