AI21 Labs policy evolution
Before/after stance changes across captured policy versions, with exact citations. If no before/after delta is available yet, AIRIN shows the latest citation-backed stance events instead.
worsenedhigh materialitytrainingAug 5, 2026 to Aug 5, 2026
model training worsened from low/no training claim to high/training permitted.
Stance key: model training. AIRIN emits this only when before and after stance events both carry citations.
Before · no training claim
low
“The Services use artificial-intelligence technology. As part of providing the Services, Personal Data may be processed, in whole or in part, by AI systems. Where AI21 acts as a controller (for example, for individual-developer accounts or for our own operational data), we process inputs and outputs to deliver and operate the Services, to return the outputs you request, to maintain security and prevent abuse, to provide support, and to maintain, troubleshoot, and improve the Services. Consistent with our Terms of Service, and unless otherwise agreed in writing or required to provide the Services, we do not use your inputs or outputs to train AI21’s foundational or generally available AI models. Because AI output is probabilistic and may be inaccurate or not unique across users, you are responsible for reviewing outputs before relying on them, and AI21 is not responsible for decisions you make based on outputs. Where AI21 processes inputs and outputs as a processor on behalf of a business customer, that processing is governed by our agreement with the customer and not by this Policy.”
Privacy Policy › “Inputs, Outputs, and AI Processing”
Open citationAfter · training permitted
high
“(e) use the Services to develop, train, or improve any product, service, or model that competes with the Services, or to benchmark or conduct competitive analysis without AI21’s prior written consent;”
§ 6.2
Open citationimprovedhigh materialitytrainingJul 19, 2026 to Aug 5, 2026
model training improved from high/training permitted to low/no training claim.
Stance key: model training. AIRIN emits this only when before and after stance events both carry citations.
Before · training permitted
high
“use any AI21 Technology to develop any service or product that is the same as (or substantially similar to), or otherwise competitive with, any AI21 Technology, or to otherwise train any artificial intelligence or machine learning models;”
§ 8.2
Open citationAfter · no training claim
low
“The Services use artificial-intelligence technology. As part of providing the Services, Personal Data may be processed, in whole or in part, by AI systems. Where AI21 acts as a controller (for example, for individual-developer accounts or for our own operational data), we process inputs and outputs to deliver and operate the Services, to return the outputs you request, to maintain security and prevent abuse, to provide support, and to maintain, troubleshoot, and improve the Services. Consistent with our Terms of Service, and unless otherwise agreed in writing or required to provide the Services, we do not use your inputs or outputs to train AI21’s foundational or generally available AI models. Because AI output is probabilistic and may be inaccurate or not unique across users, you are responsible for reviewing outputs before relying on them, and AI21 is not responsible for decisions you make based on outputs. Where AI21 processes inputs and outputs as a processor on behalf of a business customer, that processing is governed by our agreement with the customer and not by this Policy.”
Privacy Policy › “Inputs, Outputs, and AI Processing”
Open citationworsenedhigh materialitytrainingJul 8, 2026 to Jul 19, 2026
model training worsened from low/no training claim to high/training permitted.
Stance key: model training. AIRIN emits this only when before and after stance events both carry citations.
Before · no training claim
low
“Notices from Third Parties: Google – Our use of information received from Google Workspace APIs adheres to the Google AP I Services User Data Policy , including the Limited Use requirements. We do not use data obtained from these APIs to develop, improve, or train any machine learning or AI models beyond the user’s own personalized experience.”
§ 4.2.10 (Advertisement/marketing activities)
Open citationAfter · training permitted
high
“use any AI21 Technology to develop any service or product that is the same as (or substantially similar to), or otherwise competitive with, any AI21 Technology, or to otherwise train any artificial intelligence or machine learning models;”
§ 8.2
Open citationGenerated from live stance events. Informational only, not legal advice.