Ethical Stack.

Deepgram Voice Agent API

A WebSocket API that handles listening, thinking and speaking for interactive voice agents.

Score calculation

(2 + 6 + 2 + 1 + 3) ÷ 5 = 2.8

5 of six dimensions rated. Unrated dimensions are excluded from the average.

Full scoring method ↗
Setup details

Hosted Deepgram Voice Agent API; external LLM providers and telephony have separate terms.

What fits your needs

Hosted Deepgram Voice Agent API; external LLM providers and telephony have separate terms.

What it does

A WebSocket API that handles listening, thinking and speaking for interactive voice agents.

See the evidence

Route

Hosted single-WebSocket agent API; supports provider choices for STT, LLM and TTS.

See the evidence
Tradeoffs to expect

Terms and data

This is a hosted product. Review the applicable data-use, retention and model terms before sending recordings or production prompts.

See the evidence
How to switch

Get started

Use the official product or API documentation to select the relevant model and delivery route.

See the evidence
Practical questions

Rights and permissions

Check whether submitted recordings and transcripts are used for service improvement, how to disable it, how long content is kept, and how deletion works for both source audio and outputs.

See the evidence

Data path

Audio, prompts or transcripts are sent to the hosted service; connected LLMs, telephony and storage can have their own retention and training terms.

See the evidence

Energy practices

2/10↑

The public service documents a hosted route, but no comparable energy or clean-electricity report was found.

See the evidence

This hosted audio service profile scores disclosed deployment choices and operating controls, not measured energy savings. The reviewed product and company materials do not publish request-level electricity, an attributable data-center energy mix, or verified clean-power procurement for this service.

1 + 1 reporting + 0 electricity + 0 controls = 2/10. This grades documented practices and transparency.

Comparable per-minute or per-generation electricity, water use, lifecycle emissions and independently verified clean-power sourcing are not published in these sources.

Reporting1/3
The official documentation identifies the hosted runtime boundary. It does not quantify electricity or lifecycle impact for this route.
Electricity sourcing0/3
No attributable clean-electricity procurement or hourly matching evidence was found in the cited primary material.
Resource controls0/3
No documented user controls for reducing energy demand were found in these product sources.

Water disclosure

Water reporting has not been assessed for this choice.

Training data

—Depends on the model↑

The selected route can include several models, and the vendor has not published a complete audio-training rights audit here.

See the evidence

This assessment covers Deepgram Voice Agent API rather than claiming permission for an unnamed or changing model. The vendor’s terms explain how it may handle user-submitted prompts, recordings and outputs; those input rules don’t settle rights to the original model training recordings, compositions, lyrics or performer voices.

The serving model and release, training source ledger, composition and sound-recording permissions, performer consent or compensation, and removal outcomes remain unclear.

Your data & privacy

6/10↑

Hosted Deepgram API calls join its Model Improvement Program by default; a request opt-out excludes payloads from training and storage.

See the evidence

For hosted Deepgram API calls, Deepgram says API payloads are included in its Model Improvement Program by default. Set mip_opt_out=true on each request (or configure the account) to exclude payloads from model training and storage; operational metadata and usage logs are still kept for 90 days, with summarized usage retained after that. Voice Agent’s selected LLM and other connected processors follow separate terms, and region routing alone does not cover every data path. Self-hosted enterprise deployments use customer-controlled infrastructure; their data handling depends on the deployment configuration and agreement.

The reviewed docs do not give a single retention period for non-opted-out payloads; verify the request flag/account setting and separate Voice Agent processor terms.

Openness

2/10↑

The application, hosted API and model serving are proprietary; public technical docs cover integration and selection.

See the evidence

Official materials describe Deepgram Voice Agent API and how to use its hosted features, but do not publish its complete application or serving implementation under an open-source license. Public API documentation does not make the hosted model weights or full production stack inspectable.

Complete source code, model weights, training recipe and independent reproduction are unavailable for this hosted service.

Worker protections

1/10↑

Deepgram publishes public company and careers information, but the reviewed pages don’t show worker protections across its AI supply chain.

See the evidence

The reviewed company and product material provides some public visibility into the organization, but not independently verified evidence on annotation, moderation, contractor pay, worker representation, grievance handling or remedies. This score reflects what the vendor documents, not a finding that workers are mistreated.

1 + 0 + 0 + 0 = 1/10. Documented worker protections; missing disclosure is an evidence gap.

Contractor conditions, worker voice and remedy, supply-chain coverage and independent worker-level assurance are not established in these sources.

Pay & conditions0/3
No clear public evidence on pay or working conditions across employees and contractors was found in the reviewed company and product pages.
Worker voice & remedy0/3
No worker grievance, representation or remedy process was found in the reviewed public pages.
Evidence & follow-through0/3
No independent worker-level implementation audit was found in the reviewed public pages.
Workers rubric ↗

Military ties

3/10Broad defense ties↑

Deepgram publicly markets voice AI for defense missions and intelligence-community networks, with government procurement routes.

See the evidence

Deepgram’s government page advertises speech recognition, speech generation and voice agents for Defense, including on-premises and air-gapped mission deployments; it also describes classified Intelligence Community use and an In-Q-Tel portfolio. Its Carahsoft page says military commands can procure the service and gives a DoD contact-center example. This is a direct, broad defense and intelligence relationship at company level, not evidence that every catalog product is used in combat.

1 + 0 use restrictions + 0 enforcement & oversight + 2 relationship disclosure = 3/10. Known defense ties lower this rating; safeguards can improve it within the relationship’s cap.

Named operational customers, contract values, deployed model versions, mission outcomes, safeguards and whether any deployment supports combat or targeting are not established by these pages.

Use restrictions0/4
No public blanket restriction on defense or intelligence uses was found in the cited pages.
Enforcement & oversight0/3
The cited materials describe deployment and compliance choices but no independent military-use oversight or deployment review.
Relationship disclosure2/2
Deepgram names defense, intelligence-community deployments, In-Q-Tel and military procurement channels on its official pages.

Deepgram directly markets defense and classified intelligence uses and procurement routes; the reviewed pages don’t show weapons, targeting or combat-intelligence use. The relationship limits this grade to 3/10.

Military rubric ↗

BDS targets

Not listed↑

Deepgram was not named in the two official BNC technology and corporate target lists checked.

See the evidence

The official BNC corporate boycott and pressure guide and its No Tech campaign page do not name Deepgram among the current targets listed there. BDS uses selective, campaign-based targets; absence from these two lists is neither approval nor a broad assessment of company relationships.

This categorical check covers only the two linked BNC lists as reviewed on the date shown; local campaigns and future list updates may differ.

More from Deepgram

All sources

Side by side