Ethical Stack

GroqCloud

Inference API operator offering hosted models; customer data controls do not determine model training corpus rights.

Score calculation

(7 + 4) ÷ 2 = 5.5

2 of six dimensions rated. Unrated dimensions and BDS status are excluded.

Full scoring method ↗
Setup details

Groq operates inference service and may host models developed by others; this record concerns GroqCloud API.

Energy

↑

Groq has published company claims of lower LPU power use and proposed a Norway inference center; its later DOE MOU says the parties would explore power-performance benchmarks. These are infrastructure and comparative-efficiency disclosures, not verified electricity, emissions, or water per GroqCloud request.

Reasoning & sources

Groq reports an LPU efficiency comparison and a DOE collaboration to explore power benchmarks, while its Norway announcement described planned capacity. None supplies a verified footprint for a live GroqCloud model, route, or request. Ask for measured energy and water by serving region and workload.

No independently verified API-level power, region-specific grid mix, water, or per-token/request footprint is provided.

Training rights

↑

Groq says customer inputs and outputs are not used to train models; this applies to customer data and does not disclose rights in upstream model training corpora.

Reasoning & sources

Groq says customer inputs and outputs are not used to train models; this applies to customer data and does not disclose rights in upstream model training corpora. Select the model first and review its license plus corpus-provenance disclosures; prompt no-training terms do not clear the model’s original dataset.

Underlying model-specific training provenance and licenses remain open.

Privacy

↑

Groq documents no default retention of inference requests, up to 30-day temporary reliability/abuse logs, and customer-enabled ZDR; data for some GCP storage and persistence features is handled separately.

Reasoning & sources

Inference is not retained by default and ZDR is available, but temporary reliability/abuse logs may persist up to 30 days. Check the exact API/web surface, plan, region, retention setting, routing, and stateful features in the applicable terms before sending sensitive content.

Inference is not retained by default and ZDR is available, but temporary reliability/abuse logs may persist up to 30 days.

Confirm endpoint, persistence features, data-control settings, and region before use.

Openness

↑

Groq publishes its model list, including open-weight models, but its API is a hosted service and does not provide a portable serving stack.

Reasoning & sources

Some weights are open upstream; Groq’s API and runtime remain hosted and closed. Check whether the chosen model’s weights and license permit inspection or relocation, and whether the serving runtime moves too; API compatibility alone is not an open stack.

Some weights are open upstream; Groq’s API and runtime remain hosted and closed.

Model availability changes; service access does not establish rights to weights, training data, or self-hosting for each model.

Worker treatment

↑

Groq's public careers link lists openings, and its service agreement defines GroqCloud and AI Model Services, but neither source reports conditions for data workers, contractors, or data-center supply-chain workers.

Reasoning & sources

Groq's public hiring and service-contract pages identify its workforce and covered API services, but disclose no pay, exposure safeguards, or independent outcomes for labeling and infrastructure contractors. Request current vendor-level worker data and audits; a product agreement is not evidence of labor protections.

Worker pay, annotation locations, exposure controls, contractor audits, and remediation outcomes were not established.

Government & military

↑

Groq's Carahsoft agreement made GroqCloud inference available through federal procurement channels including Army ITES-SW2; a later DOE MOU concerns scientific discovery, energy-efficient inference, and Genesis Mission exploration. These establish public-sector access and a federal research relationship, not a named military model deployment.

Reasoning & sources

Groq announced access through public-sector reseller and procurement vehicles, including an Army IT contract channel, and a separate DOE MOU for scientific inference work. Those are meaningful government relationships but do not establish an awarded military deployment or specific use; request the order, customer agency, and model endpoint before scoring operational use.

No named defense award, operational military endpoint, or independent use/safeguard review is established; procurement-channel access does not prove a purchase.

BDS status

↑

We did not find Groq, Inc. among the named targets in the two official BNC campaign lists. This is a limited list check, not BDS approval.

Reasoning & sources

Groq, Inc. is not a named target in the checked lists, but BDS uses selective campaigns, so absence cannot establish a clean record. Your selected model, cloud and billing route can introduce other companies into the purchase.

This review covers the named targets in the BNC consumer/pressure guide and No Tech campaign, not every BDS campaign, local group, investment or supplier. Lists and ownership can change. A company’s absence from these lists is not an endorsement or proof that it has no relevant relationships.

All sources

Side by side