Anthropic API

Direct API access to Claude models. Anthropic’s commercial data terms govern the API; model-provider relationship and underlying training corpus remain separate questions.

Score calculation

(5 + 7 + 3 + 4 + 3) ÷ 5 = 4.4

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

Full scoring method ↗

Setup details

Anthropic develops Claude and operates its API; corporate supplier/compute partnerships are included only where directly disclosed.

Energy

↑

Anthropic reports a compute agreement with Amazon for up to 5 GW supporting Claude training and deployment. This identifies infrastructure scale/partnering, not actual energy mix, water, emissions per request, or a specific API serving path.

Reasoning & sources

The company discloses very large planned compute commitments, but no current energy mix or request-level footprint; this is a limited corporate-scale claim. For a deployment, request or export the selected service and region’s usage-linked footprint when available; do not treat company-wide renewable matching as per-request impact.

The company discloses very large planned compute commitments, but no current energy mix or request-level footprint; this is a limited corporate-scale claim.

No model-, region-, or request-level footprint established.

Anthropic and Amazon expand compute collaboration ↗

Training rights

↑

Reasoning & sources

Anthropic’s legal center lists separate consumer and commercial terms; its privacy-center disclosure says commercial chats/coding sessions are not used for training absent opt-in/feedback. These customer-data rules do not establish provenance or permissions for Claude’s pretraining corpus. Select the model first and review its license plus corpus-provenance disclosures; prompt no-training terms do not clear the model’s original dataset.

Dataset-by-dataset provenance, consent, licensing, compensation, and dispute outcomes remain outside these API data terms.

Privacy

↑

Anthropic states commercial offering chats and coding sessions are not used to train models absent explicit opt-in or specified feedback; commercial terms and retention controls are separate from consumer offerings.

Reasoning & sources

Commercial data is not used for training absent opt-in/feedback and retention rules are described, but consumer/commercial terms and feature settings differ. Check the exact API/web surface, plan, region, retention setting, routing, and stateful features in the applicable terms before sending sensitive content.

Commercial data is not used for training absent opt-in/feedback and retention rules are described, but consumer/commercial terms and feature settings differ.

Retention periods and features vary by API/contract, eligible ZDR settings, and reported feedback.

Openness

↑

Anthropic publishes SDKs and model cards, but Claude is a closed hosted service with no public weights or self-hosting path in the reviewed sources.

Reasoning & sources

SDK and system-card disclosures do not make Claude weights or serving infrastructure portable. 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.

SDK and system-card disclosures do not make Claude weights or serving infrastructure portable.

System cards are selective disclosures; model weights, full training data, and complete training methods remain unavailable in the reviewed sources.

Worker treatment

↑

Anthropic’s 2025 Claude 4 system card and inbound-services agreement disclose use of paid data-labeling contractors and require crowd-work vendors to follow minimum worker-wellbeing expectations. Stanford FMTI found no public worker compensation, location, or top-source labor-practice outcomes, so the standards remain policy commitments rather than independently verified results.

Reasoning & sources

Anthropic discloses contractor data work and vendor wellbeing requirements, while Stanford FMTI found no public pay, location, or worker-outcome data for leading sources. The commitments are relevant but not independently outcome-checked; request current vendor-level evidence before treating them as worker protections.

Public vendor-wellbeing standards and contractor disclosure are meaningful, but Stanford FMTI found no compensation or worker-level outcomes for the principal data sources.

Current pay, vendor locations, worker exposure, and compliance/audit results are not published per major data source.

Government & military

↑

Anthropic reports a two-year DoD CDAO prototype agreement with a $200 million ceiling, Claude in classified Palantir mission workflows, and Claude Gov deployments on AWS for national-security agencies. In March 2026, Anthropic said DoW designated it a supply-chain risk, narrowly applying to Claude use directly under DoW contracts, and that it would support a transition for warfighters.

Reasoning & sources

The company documents classified defense use and a $200 million prototype agreement; deployments and guardrails are not independently assessed. Request the named contract and actual model/endpoint scope with enforceable use controls; do not treat ordinary government work or silence as proof of military use.

The company documents classified defense use and a $200 million prototype agreement; deployments and guardrails are not independently assessed.

These company statements do not establish the specific operational outcomes or independently assess the safeguards of deployments.

BDS status

↑

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

Reasoning & sources

Anthropic, PBC is not a named target in the checked lists, but BDS uses selective campaigns, so absence cannot establish a clean record. Anthropic names AWS as its primary training and cloud provider and reports a large compute commitment and Amazon investment. This connects its business to a BDS pressure target without making Anthropic a separately named target; the announcement does not identify the host of every Claude request.

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