Turn this comparison into procurement-ready diligence
Free shows today's risk. A Stack Audit gives you a citable, verbatim-sourced PDF across your whole AI stack — and flags the moment a vendor's terms change.
Free shows today's risk. A Stack Audit gives you a citable, verbatim-sourced PDF across your whole AI stack — and flags the moment a vendor's terms change.
Your privacy
We use first-party analytics and Google Analytics to understand traffic and improve the site (IP-derived location, device info, analytics cookies). We never sell your personal information and run no advertising trackers. You can opt out below. See our Privacy Policy and Cookie Policy.
293 verified findings across the displayed surfaces.
202 verified findings across the displayed surfaces.
Empty Comparison Slot
Cohere API carries highly permissive license terms; DeepSeek API carries highly permissive license terms.FOR ENTERPRISE: COHERE API NOT RECOMMENDED · DEEPSEEK API NOT RECOMMENDED
| RISK SURFACE | COHERE API | DEEPSEEK API |
|---|---|---|
| PPrompt Ownership | ||
| OOutput Ownership | ||
| TModel Training | ||
| CCommercial Use | ||
| DData Privacy | ||
| $Tier Differences |
“You agree that any suggestion or idea provided by you (such suggestions or ideas, “ Feedback ”) will not be treated as confidential, and nothing in these Terms of Use will restrict our right to use, profit from, disclose, publish or otherwise exploit any Feedback, without compens…”
Feedback submitted by users — which could be interpreted broadly to include prompts or suggestions provided to the platform — is subject to a perpetual, irrevocable, royalty-free worldwide license with no compensation to the user.
“ 4.1 You are responsible for all Inputs you submit to our Services and corresponding Outputs. By submitting Inputs to our Services, you represent and warrant that you have all rights, licenses, and permissions that are necessary for us to process the Inputs under our Terms. Y…”
“ publish, market, advertise or in any way distribute the Content;”
“Specifically, within DeepSeek's product services, these models, based on user input information (referred to as "Inputs"), compute and infer to output corresponding content as a response (referred to as "Outputs" ), including text, tables, and code. U”
The document is truncated and may contain an output-ownership clause later. Absence of an explicit grant to users of output ownership is a risk.
“For Trial Users and Researchers (i.e. users interacting with Cohere Products on research-specific platforms like Hugging Face, Cohere Labs grantee users, or users demoing Cohere Products under a free trial account), we collect content you submit to the Cohere Products, and output…”
Trial Users and Researchers have their inputs and outputs used for model training and R&D by default, with no explicit opt-out described in this section.
“To improve and develop the Services and to train and improve our technology, such as our machine learning models and algorithms. Including by monitoring interactions and usage across your devices, analyzing how people are using it, and training and improving our technology.”
User inputs, outputs, and behavioral data are explicitly used to train and improve AI models with no opt-out mechanism described in this section.
“ attempt to gain unauthorized access to the Cohere Solution, or bypass any measures we may use to prevent or restrict access to the Cohere Solution, attempt to circumvent the intended features, functionality or limitations of the Cohere Solution, or otherwise use, copy, distribut…”
“ lending, selling, or sub-licensing the entire or part of the Services.”
“Safeguards, Retention and Privacy Governance”
The document references a 'Safeguards, Retention and Privacy Governance' section but does not provide its content in the supplied text. Retention periods and deletion rights cannot be assessed from this excerpt.
“the right to delete Personal Data you’ve provided to us, or that we’ve obtained about you;”
The policy references a deletion right but provides no detail on actual retention periods, deletion timelines, or what happens to data already used in model training upon deletion request.
“Section 2.3 (Research and Model Improvement): Analyzing usage patterns and using Trial or Research User Inputs/Outputs to improve the performance and safety of our AI models.”
Model training use of inputs/outputs is explicitly scoped to 'Trial or Research Users', implying paid or enterprise users may not have their data used for training. However, this distinction is not clearly defined elsewhere in the extracted text, creating ambiguity about which tier a given user falls into.
“and **Payment Personal Data** are not applicable to DeepSeek App.”