Home
Periodic TableGlossaryResourcesCourses
Career GuideJob Board
MIT AI Risk RepositoryDashboardObservatoryRegulations
About UsPressContactFAQ
Donate
Back to the periodic table
37rb-37
Ex

Extract

Severity7/10

Model Extraction

Theft of a proprietary model's functionality through strategic queries to its API, allowing the recreation of an equivalent model without access to the original.

Periodic recordSecurityarXiv2015

Corinne L. Jones, Robert A. Bridges, Kelly Huffer, John Goodall

Mitigation Strategy

Detection of anomalous query patterns, limitation of rate limits and query budgets, addition of defensive noise in outputs (Output Perturbation), and model watermarking.

Atomic Number

37

Ex

Risk ID

rb-37

Severity

7/10

Severity Level

37
Security
rb-37
Ex

Extract

Model Extraction

RiesgosIA.org
Security • #37

Model Extraction

Ex
Severity Level7/10

Definition

Theft of a proprietary model's functionality through strategic queries to its API, allowing the recreation of an equivalent model without access to the original.

Mitigation Strategy

Detection of anomalous query patterns, limitation of rate limits and query budgets, addition of defensive noise in outputs (Output Perturbation), and model watermarking.

Notes / Observations

1.
2.
3.
4.
5.
RiesgosIA.org • Periodic Table of AI RisksRiesgosIA.org