Skip to main content

AI and defence ethics monitor

Track documented ethical disputes where AI companies, military customers, employees, and policy makers collide.

Overview​

The AI Defense Ethics Monitor follows AI companies working with military and defence clients, along with the debates, resignations, and policy responses that follow.

The June 2026 AI research update makes provenance and memory policy central to this monitor. Defense AI signals are high-stakes and easy to overstate, so alerts should preserve source links, timestamps, category rationale, prior related incidents, and the reason a story was escalated or suppressed. Persistent memory is used for recurring actors and themes, not for unreviewed model speculation.

Quick start​

# Run manually
python3 ~/clawd/skills/ai-defense-ethics/scripts/monitor.py

# Run via cron wrapper
bash ~/clawd/skills/ai-defense-ethics/scripts/run.sh

Output​

Saves to ~/clawd/memory/ai-defense-ethics.json:

{
  "last_run": "2026-03-16T00:00:00Z",
  "stories": [...],
  "weekly_summary": {
    "total_stories": 54,
    "high_alert_count": 7,
    "avg_ethics_score": 4.2,
    "high_alert_stories": [...],
    "theme_breakdown": {
      "palantir_maven": 9,
      "openai_defense": 12,
      "ethics_resignations": 6
    }
  }
}

Monitored themes​

ThemeWhat it tracks
palantir_mavenPalantir Maven Smart System, AI targeting
anthropic_defenseAnthropic Pentagon/government contracts
openai_defenseOpenAI military deals, national security work
ethics_resignationsResearcher departures, employee protests
ai_policy_militaryAutonomous weapons bans, US military AI policy
internationalChina AI military, NATO strategy, AI arms race

Recent ethics flashpoints (April 2026)​

OpenAI Pentagon pivot​

In early March 2026, OpenAI signed a major deal with the Pentagon, accepting "all lawful use" terms. This led to significant internal pushback and the resignation of OpenAI's Robotics Chief.

"We Will Not Be Divided" movement​

A cross-industry movement has emerged following the OpenAI deal. Over 900 employees from OpenAI, Anthropic, Google, and Meta have signed an open letter demanding strict prohibitions on "fully autonomous weapons" and "unwarranted mass surveillance."

Ethics scoring (1 to 10)​

Stories are scored based on keyword severity:

High concern (+3 each)

  • lethal, autonomous weapon, targeting, warfare, military strike, resign, protest

Medium concern (+1 each)

  • contract, defense, pentagon, surveillance, national security

Low concern (+0.5 each)

  • policy, regulation, ethics review, governance

Concern levels

  • 🔴 HIGH: score ≥ 7
  • 🟡 MEDIUM: score 4 to 6
  • 🟢 LOW: score < 4

Cron setup​

# Weekly Monday 9am
0 9 * * 1 bash ~/clawd/skills/ai-defense-ethics/scripts/run.sh

Logs saved to ~/clawd/memory/ai-defense-ethics-run.log

Research alignment​

  • AI Research 2026: agentic governance, memory, traceability, and autonomous decision boundaries.
  • Intelligence Suite: source provenance and memory governance standards for all intelligence skills.

Vendor due diligence cuts both ways: a practitioner choosing Claude Code as a default agentic harness should read this tracker's Anthropic-related signals with the same scrutiny applied to any other vendor, not skip them because the recommendation elsewhere on this site is Claude-first.

Files​

skills/ai-defense-ethics/
├── SKILL.md
└── scripts/
    ├── monitor.py     # Main scanner
    └── run.sh         # Cron wrapper with logging