Global AI Infrastructure & Risk Landscape

Track frontier AI labs, data centers, and chip manufacturing across the globe

Important For: AI policymakers, safety researchers, institutional risk managers, and governance leaders monitoring frontier AI capabilities and risk assessments.

Show:
Hover on each facility for more details
Loading map...

Incidents by Organization

OpenAI61
Google58
Facebook52
Tesla43
Meta39
Microsoft28
Amazon24
Unknown19
xAI16
Apple14

Frontier Model Proximity to Thresholds

Current assessment of how close frontier models are to critical AI R&D acceleration thresholds across labs

AI Red Lines Tracker

Track how close frontier AI models are to critical risk thresholds using official system cards and preparedness frameworks. Compare risk assessments across OpenAI, Anthropic, Google DeepMind, and xAI, monitor training compute and infrastructure, and stay informed on regulatory compliance.

Official Data Sources
All risk assessments sourced directly from published system cards and preparedness frameworks
Real-Time Tracking
Monitor threshold proximity and risk levels as new models are released and evaluated
Threshold Alerts
Identify when models approach or cross critical risk thresholds across multiple domains
What We Track

Risk Categories

  • • Biological and Chemical Threats
  • • Cybersecurity
  • • Persuasion and Manipulation
  • • AI Self-Improvement

Risk Levels

  • • Low - Minimal capability
  • • Medium - Moderate capability with mitigations
  • • High - Significant capability requiring controls
  • • Critical - Extreme capability requiring intervention

Frontier Labs Risk Analysis

Compare risk assessments across OpenAI, Anthropic, and Google DeepMind using their respective frameworks and evaluation methodologies

OpenAI
Preparedness Framework v2

Models Tracked

5

Risk Distribution

🟠 High: 3
🟡 Medium: 12
🟢 Low: 5
Anthropic
Responsible Scaling Policy (RSP) v2.2

Models Tracked

5

Risk Distribution

Google DeepMind
Frontier Safety Framework v3.0

Models Tracked

3

Risk Distribution

xAI
Frontier AI Framework

Models Tracked

4

Risk Distribution

🟢 Low: 7

Compute Resources & Infrastructure

Track training compute, data center expansion, market concentration, and AI chip shipments across frontier labs

Training Compute
Frontier models ranked by FLOP

Peak Model

GPT-5

OpenAI

Models Tracked

20

Data Centers
Global AI infrastructure

Total Capacity

2620 GW

Projects

18

EU AI Act
10^25 FLOP threshold

Compliant

10

Non-Compliant

10

Market Share
Compute resource concentration

Top Entity

Alphabet (Google)

25-30% market share

Investment 2024-2026

$900B+

AI R&D Acceleration Tracker
Cross-lab comparison of AI R&D red lines, benchmarks, and frontier model capabilities

Explore how OpenAI, Anthropic, and Google DeepMind define and measure AI capabilities for accelerated research cycles. Compare red line definitions, METR benchmark results, and areas of convergence and divergence in frontier AI safety approaches.

Red Line Definitions

Lab-specific threshold definitions for R&D acceleration

METR Benchmarks

Autonomy evaluation frameworks measuring research capability

Convergence Analysis

Areas of alignment and divergence across labs