Geopolitics & Power
Institutions are building surveillance and enforcement infrastructure faster than accountability frameworks can form — not despite governance gaps but through them.
Signals
AI models including Claude used by US military to select strike targets in Iran. 1,000+ targets processed in 24 hours. House Defense Committee does not know how AI is used in targeting. Girls school struck, killing 170+ civilians. Anthropic filed two federal lawsuits claiming First Amendment violations after Pentagon blacklisted them for ethical stance.
AI capability operates inside the decision loop while the oversight body is outside it. Ethical constraint is now treated as adversarial to state capacity. The decision chain is opaque, the accountability surface is diffuse, and the model itself cannot be deposed.
Age-verification systems deployed under child safety laws build comprehensive adult identity databases. Technical alternatives that verify age without full identity exist (EU EUDI wallet) but are not adopted. Australian case shows 70% of under-16s retained access despite restrictions.
Think of the children is functioning as a legitimacy mechanism for surveillance infrastructure. The systems over-collect by design. The child safety mandate is the Trojan horse; the surveillance database is the payload.
US immigration agents wearing Meta AI-enabled glasses during field operations, running real-time identification. No legal framework governs data collection, retention, or access. Deployment preceded formal policy authorization.
Capability deployment ahead of governance. Technology deployed by enforcement agency, legal frameworks dont exist to constrain it, and by the time oversight is established the data and behavioral norms are entrenched.
Bluesky CEO Jay Graber stepped down. Toni Schneider (former Automattic CEO, True Ventures partner/investor) takes interim CEO. User retention sharply declined. Few independent AT Protocol server implementations exist despite decentralization claims.
Decentralized in protocol but centralized in practice. VC investor pressure for growth metrics is now the dominant governance force. The platform is caught: users attracted by anti-institutional promise, operators answering to institutional capital.
Redox OS adopted strict no-LLM contribution policy. AI-generated code immediately closed; circumvention results in permanent bans. Stated reason: asymmetric review burden — AI eliminates the proof of effort signal.
The cost of producing something now falls below the cost of evaluating it. The traditional open source model was built on labor-for-legitimacy exchange. AI contributions break the signal without breaking the form.
Control Surfaces
| Lever | Status | Change | Evidence |
|---|---|---|---|
| Congressional oversight of military AI | Absent | Degraded — committee doesnt know whats deployed | House Defense Committee disclosure |
| Federal court as AI ethics backstop | Untested | New — Anthropic lawsuit creates first major test | Two federal suits filed Mar 9 |
| Child safety as surveillance mandate | Accelerating | Expanded — facial recognition normalized in compliance | CNBC report |
| Open source governance norms | Fragmenting | New policy layer — no-LLM policies spreading | Redox OS |
| Decentralized platform governance | Collapsing toward VC norms | Deteriorated — protocol-first founder removed | Bluesky transition |
Watchlist
- ConfirmationSecond major OSS project adopting no-LLM contribution policies within 30 days
- InvalidationFederal courts dismiss Anthropic suits without engaging First Amendment merits
- ObservableWhether Iran AI targeting story produces congressional subpoena or formal inquiry within 2 weeks
Falsifiers
- Congressional oversight produces binding constraints on military AI before conflict ends
- Federal courts grant injunctions establishing precedent protecting AI companies ethical stances
- Major age-verification provider adopts EU EUDI standard
Key Unknowns
- What percentage of AI-assisted targeting decisions were reviewed by humans before execution
- Whether Anthropic First Amendment claims have viable legal theory
- Whether Redox OS no-LLM policy can be enforced at scale
Noise Filter
- Air strikes alone will not defeat Iran— Strategic military analysis, not institutional governance
- Lebanon seeks direct peace talks with Israel— Diplomatic process, no institutional structure signal
Get The Signal daily
Cross-domain structural analysis, delivered every morning.