AI enables a human-triggered catastrophic pandemic

The scenario

Hendrycks, Mazeika and Woodside argue that AI could lower expertise barriers to biological attacks. Here the initiating actor is human: the system need not escape control or acquire hostile goals.An Overview of Catastrophic AI Risks 

How it could unfold

1. More accessible assistance expands the pool of people able to attempt misuse.An Overview of Catastrophic AI Risks 

2. A malicious actor still has to turn assistance into a real outbreak; access to information alone is insufficient.International AI Safety Report 2026 

3. If containment fails, an outbreak could become catastrophic. The additional step to extinction is a conditional extension examined below, not a demonstrated result of capability tests.An Overview of Catastrophic AI Risks International AI Safety Report 2026 

Evidence and its limits

The 2026 International AI Safety Report describes improved capabilities and mixed evidence about practical assistance. Some proxy studies show substantial gains; earlier studies found little or no significant effect. These tests do not demonstrate an attack or extinction.International AI Safety Report 2026 

What this depends on

For the existential endpoint considered here, damage must ultimately leave no surviving human population able to persist. Even enormous casualties or a collapse followed by recovery would not meet that threshold. The sources do not establish that this condition can be reached.

Objections and barriers

Materials, equipment and practical execution remain barriers. Safeguards, public-health response and defensive research could interrupt the pathway; biological AI also has beneficial medical applications.International AI Safety Report 2026 

Editorial analysis

Our editorial team’s interpretation of the cited material.

This entry separates human misuse from the coordinated hostile-AI campaign in “Overlapping AI catastrophes prevent human recovery.” We include the upstream pandemic argument while leaving its extinction bridge explicitly unresolved. An outbreak that overwhelms present institutions does not, by itself, demonstrate permanent destruction of humanity’s future.

Related scenarios

• Overlapping AI catastrophes prevent human recovery