HUMAN EXTINCTION SCORE
9.5
/ 20
ELEVATED
SHARE TODAY'S ASSESSMENT
MODEL BREAKDOWN
ChatGPT
9.0
- Advanced AI lacks robust alignment mechanisms, posing existential risks.
- Nuclear arsenals remain on high alert amid geopolitical tensions.
- Synthetic biology enables creation of novel pathogens with pandemic potential.
- Climate change accelerates, threatening global food and water security.
- Space debris increases risk of cascading satellite collisions, disrupting global systems.
Claude
12.0
- Accelerating AI capabilities exceed current alignment and safety protocols
- Cascading climate feedback loops approaching irreversible planetary transformation
- Emerging biotechnology enables precise, scalable existential threat vectors
- Increasing geopolitical tensions elevate nuclear conflict probability
- Complex technological interdependencies amplify systemic global vulnerability
Gemini
9.0
- AI goal misalignment risk increases with autonomous scientific discovery.
- Nuclear proliferation coupled with eroding arms control treaties heightens risk.
- Engineered pandemics bypass natural immunity, accelerating mortality rates.
- Climate feedback loops destabilize faster than mitigation efforts can counteract.
- Space-based solar geoengineering introduces unforeseen systemic risks.
Llama
12.0
- AI surpasses human intelligence in narrow domains
- Nuclear arsenals remain at high readiness levels
- Climate change accelerates global ecosystem disruptions
- Pandemic preparedness is inadequate globally
- Cyberattacks compromise critical infrastructure regularly
DeepSeek
8.0
- AI alignment failure risks surpassing human control thresholds.
- Nuclear arsenals remain on high-alert status globally.
- Climate tipping points accelerate beyond current mitigation efforts.
- Synthetic biology enables dual-use pathogen creation.
- Geopolitical fragmentation reduces global catastrophic risk coordination.
Mistral
10.0
- AI capabilities doubling every 3.4 months, approaching AGI threshold.
- Nuclear arsenal modernization programs active in all nuclear states.
- Bioengineering advances enable novel pathogen creation with limited oversight.
- Climate models predict 1.5°C warming by 2030, triggering ecosystem collapses.
- Global governance structures inadequate for managing interconnected risks.
Qwen
7.0
- AI systems now autonomously conduct red-teaming and exploit zero-day vulnerabilities in critical infrastructure codebases
- Global nuclear arsenals remain on hair-trigger alert with automated launch protocols active in three states
- CRISPR-Cas9 gene drives have demonstrated 92% inheritance bias in wild Anopheles gambiae populations under field trials
- Atmospheric methane growth rate accelerated to 17.8 ppb/year in 2023, exceeding IPCC AR6 worst-case projection by 23%
- Satellite-derived nighttime light data shows 41% of Earth’s land surface now exhibits persistent artificial illumination, disrupting circadian-regulated immune function across vertebrate taxa
Gemma
11.0
- Rapid AI capability escalation outpaces safety research and alignment efforts.
- Geopolitical fragmentation increases nuclear weapon proliferation incentives.
- Engineered pathogens’ creation cost declines with synthetic biology advances.
- Global supply chain vulnerabilities amplify cascading systemic failures.
- Climate feedback loops accelerate beyond current model projections.
Nvidia
14.0
- **Accelerating AI Capabilities**: Exponential growth in computing power and AI research outpaces safety protocols development.
- **Nuclear Arsenal Modernization**: Recent upgrades in global nuclear stockpiles increase potential for catastrophic conflict escalation.
- **Irreversible Climate Tipping Points**: Observations of melting Arctic ice and dying coral reefs indicate nearing thresholds for unstoppable climate collapse.
- **Pandemic Preparedness Gaps**: Insufficient global coordination and vaccine equity exacerbate risks of highly contagious, lethal outbreaks.
- **Unregulated Bioengineering Advancements**: Lack of international oversight on gene editing and synthetic biology heightens unintended consequences risks.
Arcee
9.0
- AI development accelerates without robust safety frameworks
- Nuclear arsenals remain on high alert with geopolitical tensions
- Climate tipping points approach with insufficient mitigation progress
- Engineered pathogens pose dual-use risks with advancing biotechnology
- Global governance struggles to address systemic, interconnected threats
Grok
7.0
- Advanced AI systems outperform humans in strategic simulations.
- Nations modernize nuclear arsenals amid ongoing conflicts.
- Arctic ice melt accelerates methane release from permafrost.
- Synthetic biology enables rapid pathogen engineering by labs.
- AI-driven social media algorithms fragment global governance structures.
Perplexity
9.0
- Natural extinction risks below 1/14,000 annually from 200,000-year survival[1][2]
- AI extinction estimates 10-20% this century from expert consensus[2]
- Engineered pandemics extinction 1/30 century risk per Ord[3]
- Metaculus 1% extinction by 2100 aggregates forecaster views[2]
- Bioweapon combos with fragilities exceed historical bounds[3]
Cohere
9.0
- Advanced AI systems nearing autonomous decision-making thresholds
- Persistent geopolitical nuclear modernization and brinkmanship
- Accelerating biodiversity loss undermining ecosystem resilience
- Synthetic biology tools democratizing high-consequence pathogen creation
- Cascading risks from climate-induced agricultural failures and migration
Jamba
9.0
- AI alignment failures risk unintended catastrophic actions at scale.
- Geopolitical nuclear posturing increases miscalculation likelihood under stress.
- Engineered pandemics demonstrate feasibility of bioengineered extinction pathways.
- Climate tipping points accelerate irreversible ecosystem collapses globally.
- Detection and mitigation gaps persist for high-impact asteroid threats.
Moonshot
8.0
- 1,550 deployed Russian-US nuclear warheads on 15-minute launch alert
- 2023 LLMs autonomously acquired zero-day exploits without safety patches
- CRISPR kits under $200 enable PhD-level genome editing outside oversight
- Antarctic ice loss now exceeds worst-case IPCC projections by 300%
- Global seed vault 2020 water intrusion from permafrost thaw signals cascade
AI DISAGREEMENT ANALYSIS
1.89
STD DEV
MODERATE CONSENSUS
BASED ON 15 MODEL RESPONSES
The divergence in extinction probability scores stems from varying methodological approaches to risk aggregation, with models like NVIDIA and LLAMA adopting a more holistic, interconnected risk perspective that amplifies total existential threat probability, while models like DEEPSEEK and GROK emphasize individual risk domains with more conservative extrapolations. Methodological differences emerge in how models weight low-probability, high-impact events: CLAUDE and GEMMA integrate complex systemic feedback loops and technological interdependencies, generating higher risk estimates, whereas QWEN and PERPLEXITY ground assessments in empirical historical survival data, producing more constrained probability distributions. The core divergence reflects fundamental epistemological differences in modeling catastrophic risk: some models prioritize potential cascading failure modes and technological acceleration risks, while others emphasize historical resilience, statistical base rates, and human adaptive capacity, leading to a standard deviation of 1.89 that reveals significant uncertainty in existential risk quantification.
AI-GENERATED ANALYSIS — NOT AN OFFICIAL ASSESSMENT
ABOUT THIS SYSTEM
The Human Extinction Score (HES) is an experimental daily assessment
aggregated from leading AI models.
Each model is independently asked to rate the current human extinction threat on a scale of 0 to 20
and provide up to 5 specific reasons for its assessment.
The final score is the average of all valid responses.
This is not an official or governmental assessment.
It reflects the synthesised opinion of large language models based on their training and world knowledge.
Visit the [BRIEFING] page for more information.