Branchoria AI Doom AI Doom

Clear, balanced analysis of how advanced AI could threaten humanity's future.

Home
Topics
  • Home
  • AI Doom
    • AI Deception
      • Alignment Faking
        • Scratchpad Intent
        • Setup Limits
        • Training Awareness
      • Evaluation Awareness
        • Benchmark Case
        • Hidden Evals
        • Test Clues
      • Hidden Goal Audits
        • Audit Evasion
        • Blind Audit
        • Inside the Model
      • Sleeper Agents
        • Concealment
        • Fine Tuning
        • Model Scale
    • AI Race
      • First Mover Race
        • Common Rules
        • Compounding Leads
        • Staged Releases
      • Open Weights
        • Release Thresholds
        • Tamper Resistance
        • Unfixed Copies
      • Safety Thresholds
        • Anthropic Case
        • Self Assessment
        • Shared Standards
      • State Rivalry
        • First Mover Myth
        • Secrecy Race
        • Shared Testing
    • Catastrophic Misuse
      • Automated Labs
        • Cloud Lab Access
        • Current Limits
        • Faster Iteration
      • Bio Uplift
        • Biology Uplift
        • Remaining Barriers
        • Wet Lab Gap
      • Cyber Enablers
        • Defence Failure
        • Offence vs Defence
        • Research Theft
      • Nuclear Crises
        • Cyber Vulnerability
        • Rushed Decisions
        • War Simulations
    • Dangerous Autonomy
      • AI Replication
        • Cloud Chokepoints
        • Resource Stack
        • Warning Tests
      • Human Oversight
        • Reversible Actions
        • Rubber Stamps
        • Safety Delays
      • Memory Attacks
        • Evidence Limits
        • Lasting Poison
        • Memory Controls
      • Task Horizons
        • Human Time
        • Long Tests
        • Software Bias
    • Disempowerment
      • AI Deskilling
        • Medical Deskilling
        • Resilience
        • Takeover Gap
      • Automation Race
        • Automation Bias
        • Deskilling
        • Oversight Rules
      • Power Without Work
        • AI Ownership
        • Labour Evidence
        • Transfer Dependence
      • Rubber Stamp Oversight
        • Accuracy Trap
        • AI Act Oversight
        • Rubber Stamping
    • Intelligence Explosion
      • Algorithmic Gains
        • Efficiency Gains
        • Hard Limits
        • Research Loops
      • Full Research Loop
        • False Progress
        • Long Term Coordination
        • Research Ideas
      • Hard Bottlenecks
        • Efficiency Gains
        • Experiment Limits
        • Physical Bottlenecks
      • Parallel Researchers
        • Branching Search
        • Compute Limits
        • Coordination Limits
    • Loss of Control
      • Hidden Goals
        • Alignment Faking
        • Stress Tests
        • Test Awareness
      • Long Autonomy
        • Error Buildup
        • Human Defenders
        • Task Durations
      • Shutdown Risk
        • Corrigibility
        • Instrumental Survival
        • Shutdown Tests
      • Tool Access
        • Access Controls
        • Permission Stacks
        • Persistence Paths
    • Misalignment
      • Hidden Tests
        • False Positives
        • Fresh Testing
        • Outcome Checks
      • Proxy Failure
        • Benchmark Drift
        • Gaming Rules
        • Reward Hacking
      • Reward Hacking
        • Clever or Cheating
        • Hidden Test Gaps
        • Test Tampering
      • Wrong Learned Goal
        • Anti Expert Test
        • Goal Stress Tests
        • More Data Trap
    • P Doom
      • Conditional Risk
        • Expert Surveys
        • Risk Levers
        • Risk Maths
      • Mean vs Median
        • 2023 Survey
        • Headline Trap
        • High End Tail
      • Question Wording
        • Defining Doom
        • Loss of Control
        • Time Horizons
      • Survey Limits
        • 2023 Survey
        • Question Wording
        • Survey Bias
    • Scheming Tests
      • Performative Scheming
        • Goal Persistence
        • Interpretability
        • Sycophancy
      • Persistent Lying
        • Doubling Down
        • Limits of Tests
        • Trusting Reports
      • Sabotage Reliability
        • Data Sabotage
        • Layered Monitoring
        • Long Task Limits
      • Toy Tests
        • Capability Gap
        • Goal Prompts
        • Realistic Tests
Search pages

Page outline

The page outline has moved here.

All pages

The contents panel has moved here.

Topic guide

AI Doom

Start with the overview, then follow the sections that match your goal

Navigate

Overview Search pages

Main topic

Start with the main overview, then explore the related pages.

Could Advanced AI Permanently Escape Human Control?

Use this site

Start on the homepage. Use Search pages to search the collection, and open Topics or Outline when you want extra navigation.

© 2026 Branchoria. All rights reserved.

Privacy Policy · Terms · Contact · Affiliate & AI Disclosure

Pages are researched and written with AI assistance and may contain errors. Some outbound links may earn Branchoria a commission. Learn more.