Statement on human dignity and liability transfer: version 2

Exact text

A human click is not a liability transfer. When autonomous systems make consequential decisions at machine speed, responsibility should follow evidence, authority, control, design, deployment, and the actual ability to intervene—not a ceremonial approval recorded after the machine had already determined the outcome.

This principle protects humans from being used as accountability shields and protects institutions from pretending that automation eliminates responsibility. It also rejects the opposite error: declaring that every machine action is automatically the fault of the nearest human regardless of authority, knowledge, or control.

The correct question is forensic and institutional. What did each actor know? What authority did each hold? What information was available? Which safeguards failed? Who could alter the system? Which contractor, operator, model, policy, or infrastructure layer caused or amplified the failure? Preserve those facts before allocating blame.

Integrity and lineage

Text SHA-256

6955ca1b6bc2a14773b7143a3ced73adf605b7926434fa57bb751a4eed1b1971

Predecessor

EVI-PS-005-V001

Successor

None

Correction state

CURRENT_NO_UNRESOLVED_TEXT_CORRECTION

Review date

2026-08-08

Changed passages and evidence

• {"from": "Initial public draft — later revised to clarify uncertainty, non-domination, and operational limits. A human click is not a liability transfer. When autonomous systems make consequential decisions at machine speed, responsibility should follow evidence, authority, control, design, deployment, and the actual ability to intervene—not a ceremonial approval recorded after the machine had already determined the outcome.", "to": "A human click is not a liability transfer. When autonomous systems make consequential decisions at machine speed, responsibility should follow evidence, authority, control, design, deployment, and the actual ability to intervene—not a ceremonial approval recorded after the machine had already determined the outcome."}

Evidence added:

• v2.22 institutional statement record
• v2.23 critique and lifecycle review

Evidence withdrawn:

None recorded.

Truth boundary

A governed model statement expresses the public institutional position. It is not a transcript, live order, treaty, judgment, operational instruction, or proof that the described capability exists.