Misuse catastrophe
A major biological or cyber attack is traced to AI, and governments lock down access to capable models.
Cuts off 11 boxes, 6 of 53 stages · slows 30 · Show on the map →
Hits directly
Signs
In June 2026 the US briefly restricted two frontier models on national-security grounds over cyber capability. OpenAI said in 2025 its models were nearing "High" capability in biology, and a state-linked group has already used an agent for most of a hacking campaign.
What would soften it
Biosecurity safeguards and AI cyber defence that keep pace with capability.
Cut off 11
Slowed 30
- White-collar hiring slowdown
- Cognitive abundance
- AI research tools in investing
- Frontier lab revenue
- On-device models
- Sycophancy
- AI-enabled persuasion
- Knowledge-rich agents
- AI-native application revenue
- Continual-learning agents
- Firms shrink to a few operators
- Public AI scepticism
- Software that maintains itself
- Agent-native organisations
- AI-run capital allocation
- Drop-in remote worker
- Scalable oversight
- Agent economies
- Agent-run discovery
- Agents designing agents
- Alignment
- Civilisation-scale knowledge
- Explosive economic growth
- Internal deployment oversight
- Long-horizon value stability
- Superintelligence
- Human-agent symbiosis
- Long-latency civilisation
- Choosing the objective
- Extended mind