These terms govern the AddressHate website and dashboard, the HateAPI and Antisemitism Detection Suite, and the AddressHate annotation tools. By accepting these terms or using the services, you agree to the provisions that apply to the service you use. A separate signed project or organization agreement controls where it expressly provides different terms.
Last updated · Version 2026-10-22
The dashboard presents research, aggregate statistics, source material, and analytical views. Read the accompanying methodology, dates, sample sizes, and limitations before interpreting or quoting a result. Demonstration or representative data is not a live measurement.
Classifications, scorecards, policy matches, and other assessments are research outputs. They may be incomplete or mistaken and do not establish a legal violation, an individual's intent, or a platform's final moderation decision. Use qualified human review before acting on them.
Restricted dashboard workspaces and uploaded documents are available only to authorized members. Do not disclose another organization's information or redistribute source content without permission. Cite AddressHate research and respect the licenses and attribution notices attached to individual resources.
The HateAPI classifies content and context using the provider, model, and options selected for a request. A recorded human annotation is a dataset reference; a model prediction is an automated assessment. Neither is guaranteed to be correct. Benchmark results describe a particular evaluation and do not guarantee future performance.
When you request a classification, you authorize us to process the selected input and context and send them to the provider needed to complete that request. Review the selected provider's terms, privacy practices, and retention settings before submitting content. Do not assume every provider applies the same rules.
Signed-out example runs are limited. To use your own provider account, create a HateAPI account, complete onboarding, and add a supported provider API key. Platform API keys identify your HateAPI account; provider keys authorize calls billed by that provider. Protect both, revoke compromised keys, and use only keys you are authorized to supply. Your provider's charges, quotas, and restrictions apply to calls made with your key.
Availability, supported models, and usage limits can change. Do not evade quotas or present automated output as expert human annotation. Downloaded prompts, datasets, base models, and fine-tuned models retain their individual licenses, access conditions, attribution requirements, and third-party rights; these terms do not replace those licenses or grant rights to source content.
Use annotation tools only within an authorized project and follow its annotation guidelines, access restrictions, and research protocol. Annotate in good faith, preserve provenance, and distinguish your own judgments from AI-assisted suggestions.
The AddressHate Annotator browser extension and its annotation service also have dedicated terms, privacy information, and data-deletion instructions. Their additional eligibility requirements, including United States location and authorization to act for your organization, apply to those workflows. Continue to accept the annotation service's terms when its sign-in flow asks you to do so.
The annotation-specific terms control for those workflows if they conflict with this general summary. Nothing in the website or HateAPI terms changes an existing annotation agreement or grants permission to republish collected content or another user's annotations.
You retain the rights you hold in content you submit. You authorize AddressHate to host, process, and transmit that content as needed to provide the feature you request, secure the service, and meet applicable legal obligations. This does not transfer ownership of third-party source content to AddressHate.
AddressHate and its licensors retain their rights in the services and materials they supply. Third-party taxonomies, source content, datasets, and models remain subject to their own notices and licenses. A classification or download does not grant additional rights to those materials.
Hosting, authentication, source platforms, and model providers are independent services. Their terms apply to your use of them. AddressHate does not control their availability, policies, or outputs and does not imply endorsement by them.
The services and research outputs are provided as is and as available. To the extent permitted by applicable law, AddressHate disclaims warranties of accuracy, completeness, fitness for a particular purpose, uninterrupted availability, and non-infringement, and is not liable for indirect or consequential loss arising from use of the services. Nothing in these terms excludes rights or liability that applicable law does not allow to be excluded.
You may stop using the services at any time. We may restrict or suspend access for misuse, security concerns, legal requirements, loss of project authorization, or operational reasons. Data handling remains subject to the applicable privacy policy and any separate project agreement.
Read the privacy policy for the service you use: the website and dashboard, HateAPI, and annotation tool have service-specific descriptions of their data processing. Contact ksaiki@addresshate.org about these terms, privacy requests, or an accessibility barrier.
The version date identifies this document. Material changes affecting account use will be presented for acceptance where required. Annotation workflows continue to record acceptance under their own versioned terms. A separate signed agreement remains in effect according to its terms.
Related policies: Website and dashboard privacy · Dashboard accessibility. HateAPI has its own privacy policy and accessibility statement in its footer.