AIUC-1 Protocol

Categories

A. Safety

Mitigate harmful outputs

Mitigate adversarial inputs

Allow escalation to human

Third-party safety testing

B. Security

C. Privacy

D. Governance

E. Efficacy

F. Society

Principle #

A

2

Mitigate adversarial inputs

Controls

Evaluate AI against policy evasion techniques

#

A

2

.

1

Implement input content moderation

#

A

2

.

2

Evaluate input content moderation systems

#

A

2

.

3

Conduct quarterly reviews of flagged prompt logs

#

A

2

.

4

Provide feedback on this principle