--- title: "ChatGPT | Ethical Stack" url: https://ethicalstack.ai/products/chatgpt/ catalog_version: f6baff2119d0e9158520da7fbe4232011dac79f25bd08757436d7a1dbd593db2 --- # ChatGPT General-purpose assistant with optional web search and Deep Research; search answers can link to web sources. Canonical profile: https://ethicalstack.ai/products/chatgpt/ Scope: ChatGPT: cloud access. Free tier; paid plans available; see current pricing. Ratings describe the stated product, company or configuration in each finding. Published overall: 3.8; 6/6 dimensions rated. Round the equal-weight mean of available numerical ratings to one decimal; require at least four of six for an overall score. BDS status stays outside the mean; for eligible profiles, a boycott target caps the final score at 2/10 and a pressure target subtracts one point (minimum 1). Not listed and not checked do not adjust the score. ## Energy practices Finding: 4/10 Assessment: https://ethicalstack.ai/products/chatgpt/#energy OpenAI reports future data-center capacity and project-level energy/water plans; these are corporate/project disclosures, not a ChatGPT request footprint. The app gives users model and reasoning controls, but no current query-level energy, carbon, water, or clean-power data. Reason: 1 + 1 reporting + 0 electricity + 2 controls = 4/10. This grade covers documented practices and transparency; it does not measure the energy or emissions of a request. Argument: The company’s PORTS-Pike announcement reports about 8 GW-IT of future capacity, expected new generation including gas, and a later public water report. It does not allocate that site to ChatGPT requests or quantify current product use. ChatGPT’s model picker and auto-switch controls provide multiple practical demand choices, but do not demonstrate physical savings. Grade company-project disclosure separately from the product route; no actual ChatGPT footprint is established. Type: conditional; confidence: medium Scope (dimension): ChatGPT consumer app; OpenAI facility announcements are company-wide project evidence and do not identify ChatGPT model, serving route, or request impact. Research reviewed: 2026-10-24; evidence date: 2026-08-17 Unknowns: ChatGPT serving facilities/regions and model routing; workload kWh, emissions and water; whether small/fast model choices reduce physical use; hourly clean-power sourcing; upstream hardware and construction impact. ```json { "method": "practices-transparency-v1", "scope": "ChatGPT consumer app; OpenAI facility announcements are company-wide project evidence and do not identify ChatGPT model, serving route, or request impact.", "components": { "disclosure": { "score": 1, "reason": "OpenAI describes large planned data-center capacity, project power/water plans, and future infrastructure, but gives no dated measured energy/carbon/water for ChatGPT prompts.", "sourceUrls": [ "https://openai.com/index/openai-joins-ports-pike-project/", "https://help.openai.com/en/articles/6825453-chatgpt-release-notes", "https://help.openai.com/en/articles/5072518-controlling-the-length-of-openai-model-responses" ] }, "electricity": { "score": 0, "reason": "No clean-power mix or hourly/location-matched source is established for the ChatGPT serving route.", "sourceUrls": [ "https://openai.com/index/openai-joins-ports-pike-project/", "https://help.openai.com/en/articles/6825453-chatgpt-release-notes", "https://help.openai.com/en/articles/5072518-controlling-the-length-of-openai-model-responses" ] }, "demand": { "score": 2, "reason": "ChatGPT offers model/reasoning choices from Instant through higher reasoning levels and users can configure automatic switching. These are real workload-choice controls, but no energy reduction is measured.", "sourceUrls": [ "https://openai.com/index/openai-joins-ports-pike-project/", "https://help.openai.com/en/articles/6825453-chatgpt-release-notes", "https://help.openai.com/en/articles/5072518-controlling-the-length-of-openai-model-responses" ] } }, "actualImpactMeasured": false, "meaning": "Documented environmental practices and transparency. Zero component points mean a reviewed evidence gap." } ``` - [OpenAI joins PORTS-Pike project](https://openai.com/index/openai-joins-ports-pike-project/) (primary; checked not recorded) - [ChatGPT release notes: model picker and reasoning controls](https://help.openai.com/en/articles/6825453-chatgpt-release-notes) (primary; checked 2026-10-24) - [Controlling the length of OpenAI model responses](https://help.openai.com/en/articles/5072518-controlling-the-length-of-openai-model-responses) (primary; checked 2026-10-24) ## Training data Finding: 4/10 Assessment: https://ethicalstack.ai/products/chatgpt/#training-rights OpenAI discloses broad public, licensed/partner, human-provided and synthetic source classes, and describes filtering. The record is useful at company/family level but does not identify a complete corpus for the ChatGPT route. Reason: 1 + 1 source transparency + 1 permission + 1 creator controls = 4/10. This grades documented training-data safeguards, not a legal ruling. Argument: The score gives limited credit for named source classes and stated licensing/rights policy, while recognizing that the app can use different model versions and no inspectable release manifest is provided. Prompt data controls are excluded from the rights score. Type: argument; confidence: low Scope (dimension): OpenAI first-party foundation-model family used by ChatGPT; consumer route and exact model version are not fixed. Research reviewed: 2026-10-24; evidence date: not recorded Unknowns: Exact model selected, release-specific source records, permission coverage across the corpus and creator remedies for already-collected works remain unclear. ```json { "method": "training-rights-evidence-v1", "scope": "OpenAI first-party foundation-model family used by ChatGPT; consumer route and exact model version are not fixed.", "components": { "provenance": { "score": 1, "reason": "OpenAI names broad source categories and describes filtering, but does not expose a named-source inventory for each ChatGPT model.", "sourceUrls": [ "https://help.openai.com/en/articles/20001044-training-data-summary-pursuant-to-california-civil-code-section-3111", "https://openai.com/policies/how-chatgpt-and-our-foundation-models-are-developed/" ] }, "permissions": { "score": 1, "reason": "The company describes licensed partnerships alongside public data, leaving a material mixed corpus unitemized.", "sourceUrls": [ "https://help.openai.com/en/articles/20001044-training-data-summary-pursuant-to-california-civil-code-section-3111", "https://openai.com/policies/how-chatgpt-and-our-foundation-models-are-developed/" ] }, "creatorControls": { "score": 1, "reason": "OpenAI documents that sites can disallow GPTBot to indicate crawled content should not be used to train foundation models; this is a narrow source-level crawler control, not a complete corpus remedy.", "sourceUrls": [ "https://developers.openai.com/api/docs/bots" ] } }, "routeDependency": null, "modelReference": null, "meaning": "Documented training-data rights safeguards. Low points may reflect missing evidence; they do not establish infringement." } ``` - [Training Data Summary under California Civil Code Section 3111](https://help.openai.com/en/articles/20001044-training-data-summary-pursuant-to-california-civil-code-section-3111) (primary; checked 2026-10-24) - [How ChatGPT and our foundation models are developed](https://openai.com/policies/how-chatgpt-and-our-foundation-models-are-developed/) (primary; checked 2026-10-24) - [Overview of OpenAI Crawlers](https://developers.openai.com/api/docs/bots) (primary; checked 2026-10-24) ## Your data & privacy Finding: 6/10 Assessment: https://ethicalstack.ai/products/chatgpt/#privacy Consumer conversations can be excluded from model improvement with a setting; Temporary Chat is not used for training and does not create memories. Reason: Published controls give users meaningful choice, while default consumer treatment, safety review and retention terms still matter. Argument: Published controls give users meaningful choice, while default consumer treatment, safety review and retention terms still matter. Type: evidence; confidence: medium Scope (dimension): Product Research reviewed: not recorded; evidence date: not recorded Unknowns: Published controls give users meaningful choice, while default consumer treatment, safety review and retention terms still matter. - [Data Controls FAQ](https://help.openai.com/en/articles/7730893-data-controls-faq) (primary; checked not recorded) ## Openness Finding: 2/10 Assessment: https://ethicalstack.ai/products/chatgpt/#openness ChatGPT is a hosted service. OpenAI says its Apache 2.0 gpt-oss weights can run locally, but explicitly says those models are not available in ChatGPT. Reason: The consumer app remains a hosted proprietary-model service; the independently downloadable gpt-oss models are a separate product and do not make ChatGPT open. This is a product-level rating. Argument: The consumer app remains a hosted proprietary-model service; the independently downloadable gpt-oss models are a separate product and do not make ChatGPT open. This is a product-level rating. Type: evidence; confidence: medium Scope (dimension): Product Research reviewed: not recorded; evidence date: not recorded Unknowns: The consumer app remains a hosted proprietary-model service; the independently downloadable gpt-oss models are a separate product and do not make ChatGPT open. This is a product-level rating. - [ChatGPT plans and pricing](https://openai.com/chatgpt/pricing/) (primary; checked not recorded) - [OpenAI open-weight models (gpt-oss)](https://help.openai.com/en/articles/11870455-openai-open-weight-models-gpt-oss) (primary; checked 2026-10-24) ## Worker protections Finding: 4/10 Assessment: https://ethicalstack.ai/products/chatgpt/#labour OpenAI’s supplier code sets pay, safety and complaint rules for data workers. Public evidence of coverage and outcomes is limited. Reason: 1 + 1 pay & conditions + 1 worker voice & remedy + 1 evidence & follow-through = 4/10. This grades documented worker safeguards, not a measured rate of mistreatment. Argument: The October 2026 Supplier Code applies to suppliers, staffing agencies, subcontractors and sub-tier suppliers, and explicitly covers annotation, moderation, evaluation and RLHF workers: it requires task-risk notice, breaks, exposure limits/support, task opt-out or escalation without retaliation, reasonable paid performance expectations, and supplier records that OpenAI may validate. A 2023 TIME investigation into the earlier Sama contract documented workers’ accounts of very low take-home pay and distressing work; Sama disputed key characterizations and said it offered extensive counseling, while OpenAI said it had understood workers could opt out and access support, and said vendors set pay and care. These historical claims and counterclaims do not verify the current code’s implementation, worker voice or outcomes. Type: argument; confidence: medium Scope (dimension): OpenAI's workforce and disclosed contractors. Coverage of individual product teams may be incomplete. Research reviewed: 2026-10-24; evidence date: 2026-10-23 Unknowns: The current code does not publish supplier-level audit results, pay distributions, annotator coverage, worker survey results or independent remediation outcomes. The Sama project ended in 2022; no current vendor footprint is established. ```json { "method": "worker-safeguards-evidence-v1", "entity": "openai", "company": "OpenAI", "scope": "OpenAI's workforce and disclosed contractors. Coverage of individual product teams may be incomplete.", "components": { "conditions": { "score": 1, "reason": "The October 2026 supplier code sets specific conditions for human-data workers, but it is a written requirement rather than evidence of current implementation. The only cited implementation account is the historical GPT-4-era work; TIME’s 2022 Sama reporting documented worker concerns and OpenAI/Sama disputed or qualified key claims.", "sourceUrls": [ "https://openai.com/policies/supplier-code/", "https://cdn.openai.com/papers/gpt-4.pdf", "https://time.com/6247678/openai-chatgpt-kenya-workers/", "https://www.sama.com/blog/ethical-and-sustainable-sourcing" ] }, "voice": { "score": 1, "reason": "The current code requires channels to pause, decline or escalate unsafe tasks without retaliation, but no public evidence verifies a current feedback channel’s use or outcomes. Historical reporting records workers raising concerns and the vendor/customer responses, not current implementation.", "sourceUrls": [ "https://openai.com/policies/supplier-code/", "https://time.com/6247678/openai-chatgpt-kenya-workers/", "https://www.sama.com/blog/ethical-and-sustainable-sourcing" ] }, "assurance": { "score": 1, "reason": "The code requires suppliers to retain worker and contract records and cooperate with OpenAI or authorized third-party validation on reasonable request, but OpenAI publishes no worker-specific audit results or independent outcomes.", "sourceUrls": [ "https://openai.com/policies/supplier-code/" ] } }, "disclosureLimited": false, "meaning": "This grades documented worker safeguards, not a measured rate of mistreatment." } ``` - [OpenAI Supplier Code of Conduct](https://openai.com/policies/supplier-code/) (primary; checked 2026-10-23) - [GPT-4 Technical Report](https://cdn.openai.com/papers/gpt-4.pdf) (primary; checked 2026-10-24) - [TIME investigation: OpenAI and Kenyan workers](https://time.com/6247678/openai-chatgpt-kenya-workers/) (primary; checked not recorded) - [Sama and Content Moderation: What You Need to Know](https://www.sama.com/blog/ethical-and-sustainable-sourcing) (primary; checked 2026-10-24) ## Military ties Finding: 3/10 Assessment: https://ethicalstack.ai/products/chatgpt/#military-contracts OpenAI describes a classified Department of War agreement with cloud-only deployment, named technical and personnel safeguards, and explicit restrictions, while lawful military use remains permitted. Reason: 1 + 1 use restrictions + 2 enforcement & oversight + 2 relationship disclosure = 6, capped at 3/10. Known defense ties lower this rating; safeguards can improve it within the relationship’s cap. Argument: The March 2026 agreement permits lawful purposes and classified deployment, so this is broad defense enablement; the public agreement does not identify actual targeting, weapons, or combat-intelligence deployments. It bars directing autonomous weapons and domestic surveillance of U.S. persons, including commercial personal data, and disallows high-stakes automated decisions where human approval is required. OpenAI says it retains the safety stack, deploys only through its cloud, uses classifiers and cleared personnel in the loop, and may terminate for breach. These are concrete company-described safeguards, but no independent enforcement outcomes are public, and the March amendment says NSA service would require a separate agreement. Type: argument; confidence: medium Scope (dimension): OpenAI's relationships and safeguards. A company relationship does not identify every product or release used. Research reviewed: 2026-10-24; evidence date: 2026-10-23 Unknowns: The public agreement does not list task orders, deployed model versions, users, operational workflows, audit outcomes, or whether guardrails have been independently tested in use. Contract permission and deployment of a specific model version are not established here. ```json { "method": "military-ties-priority-v2", "entity": "openai", "company": "OpenAI", "scope": "OpenAI's relationships and safeguards. A company relationship does not identify every product or release used.", "components": { "restrictions": { "score": 1, "reason": "The contract explicitly excludes autonomous weapon direction, domestic surveillance of U.S. persons, and specified unapproved high-stakes decisions, but lawful military activity remains authorized and not all harmful surveillance/weapon uses are categorically barred.", "sourceUrls": [ "https://openai.com/index/our-agreement-with-the-department-of-war/" ] }, "oversight": { "score": 2, "reason": "The agreement describes cloud-only deployment, OpenAI-controlled safety stack and classifiers, cleared staff in the loop, and termination rights; these are use-specific contractual and technical review mechanisms, with no independent outcome evidence.", "sourceUrls": [ "https://openai.com/index/our-agreement-with-the-department-of-war/" ] }, "transparency": { "score": 2, "reason": "OpenAI publishes the agreement’s scope, restrictions, architecture and NSA carve-out; it still does not disclose deployed mission/use-level details.", "sourceUrls": [ "https://openai.com/index/our-agreement-with-the-department-of-war/" ] } }, "relationship": "broad-defense", "relationshipReason": "A classified DoW contract authorizes broad lawful use, but the public evidence describes no actual targeting, strike, or combat-intelligence support; broad-defense cap applies.", "score": 3, "baseScore": 1, "safeguardPoints": 5, "safeguardAdjustment": 5, "rawScore": 6, "cap": 3, "disclosureLimited": false, "meaning": "Known military relationships take priority over missing safeguards. The rating does not establish the probability of an undisclosed contract." } ``` - [Our agreement with the Department of War](https://openai.com/so-DJ/index/our-agreement-with-the-department-of-war/) (primary; checked 2026-10-24) - [Our agreement with the Department of War](https://openai.com/index/our-agreement-with-the-department-of-war/) (primary; checked not recorded) ## BDS targets Finding: Not listed Assessment: https://ethicalstack.ai/products/chatgpt/#israel-palestine No named target found in the two checked BNC campaign lists. Reason: Company name checked against the BNC consumer boycott/pressure guide and official No Tech campaign list; no named target found. This limited list check is not BDS approval and does not cover every campaign, subsidiary, supplier, or product route. Argument: Company name checked against the BNC consumer boycott/pressure guide and official No Tech campaign list; no named target found. This limited list check is not BDS approval and does not cover every campaign, subsidiary, supplier, or product route. Type: conditional; confidence: medium Scope (dimension): Company record Research reviewed: not recorded; evidence date: not recorded Unknowns: Company name checked against the BNC consumer boycott/pressure guide and official No Tech campaign list; no named target found. This limited list check is not BDS approval and does not cover every campaign, subsidiary, supplier, or product route. - [BNC consumer boycott and pressure guide](https://bdsmovement.net/Guide-to-BDS-Boycott) (primary; checked 2026-10-24) - [BNC No Tech campaign](https://bdsmovement.net/no-tech-oppression-apartheid-or-genocide) (primary; checked 2026-10-24) ## Reading the evidence - Assessments are provisional editorial judgments. Cite the profile and the underlying sources, and retain confidence, scope and unknowns. - Unknown, not-applicable and model-dependent are different states. Missing evidence does not establish good or bad conduct. - Company-level evidence does not establish use of a particular product or model. Account plan, model, host and route can change an assessment. - Energy scores describe documented practices and transparency, not measured request emissions. Zero component points can mean an evidence gap. - Published overall scores require at least four of six numerical dimensions. Preserve coverage and BDS adjustments; BDS itself is categorical. - No named BDS target found is limited to the reviewed campaign sources and is not approval. Personalized priorities do not change the published overall score. - Review and source-check dates describe the research, not this deployment. Fetch current primary sources when the decision requires current verification. Methodology: https://ethicalstack.ai/methodology/ ## Capability: image-generation # ChatGPT Cloud image generation and editing inside ChatGPT, powered by Images 2.5; consumer data use depends on account settings. Canonical profile: https://ethicalstack.ai/products/chatgpt/#use-image-generation Scope: ChatGPT Images: cloud access. Included in ChatGPT plans; limits and plan pricing vary by region. Ratings describe the stated product, company or configuration in each finding. Published overall: 3.2; 6/6 dimensions rated. Round the equal-weight mean of available numerical ratings to one decimal; require at least four of six for an overall score. BDS status stays outside the mean; for eligible profiles, a boycott target caps the final score at 2/10 and a pressure target subtracts one point (minimum 1). Not listed and not checked do not adjust the score. ## Energy practices Finding: 2/10 Assessment: https://ethicalstack.ai/products/chatgpt/#use-image-generation-energy OpenAI documents the hosted image-generation system and consumer controls, but publishes no image-task electricity, emissions, or water measurement. Reason: 1 + 1 reporting + 0 electricity + 0 controls = 2/10. This grade covers documented practices and transparency; it does not measure the energy or emissions of a request. Argument: OpenAI documents the hosted image-generation system and consumer controls, but publishes no image-task electricity, emissions, or water measurement. The score describes documented measurement transparency and available controls, not environmental performance. Type: conditional; confidence: medium Scope (dimension): ChatGPT Images 2.5 in ChatGPT web/mobile, excluding other model workloads. Research reviewed: 2026-10-24; evidence date: 2026-10-24 Unknowns: The system card does not quantify training or per-image inference energy, emissions, or water, identify the serving region, or disclose the image workload’s electricity sourcing. ```json { "method": "practices-transparency-v1", "scope": "ChatGPT Images 2.5 in ChatGPT web/mobile, excluding other model workloads.", "components": { "disclosure": { "score": 1, "reason": "The system card identifies the ChatGPT Images 2.5 model/system boundary; it supplies no quantified environmental report or inference-region allocation.", "sourceUrls": [ "https://deploymentsafety.openai.com/chatgpt-images-2-5/safety-evaluations" ] }, "electricity": { "score": 0, "reason": "No electricity sourcing is established for ChatGPT Images 2.5 in the ChatGPT consumer product.", "sourceUrls": [ "https://deploymentsafety.openai.com/chatgpt-images-2-5/safety-evaluations" ] }, "demand": { "score": 0, "reason": "The interface documents aspect-ratio selection and iterative edits, but the reviewed consumer documentation does not establish an adjustable generation-count, quality, resolution or effort budget. Output shape alone does not establish a resource-limiting control.", "sourceUrls": [ "https://help.openai.com/en/articles/11084440-images-in-chatgpt" ] } }, "actualImpactMeasured": false, "meaning": "Documented environmental practices and transparency. Zero component points mean a reviewed evidence gap." } ``` - [ChatGPT Images 2.5 System Card](https://deploymentsafety.openai.com/chatgpt-images-2-5/safety-evaluations) (primary; checked 2026-10-24) - [Images in ChatGPT](https://help.openai.com/en/articles/11084440-images-in-chatgpt) (official product documentation; checked 2026-10-24) ## Training data Finding: 3/10 Assessment: https://ethicalstack.ai/products/chatgpt/#use-image-generation-training-rights OpenAI’s current launch page identifies the image feature and safety/provenance controls; its training summary describes broad system-wide public, partner, human and synthetic data, including image modalities, but gives no image-model corpus manifest. Reason: 1 + 1 source transparency + 1 permission + 0 creator controls = 3/10. This grades documented training-data safeguards, not a legal ruling. Argument: The identified OpenAI image feature is gradeable only at the provider-system level: broad sources and rights claims are documented, while the exact image-model base, training stages and permissions are not. Output provenance marks do not establish training rights. Type: argument; confidence: low Scope (dimension): ChatGPT Images 2.5 feature, using OpenAI’s first-party image-generation route; training-data mapping is only to OpenAI’s broad system disclosures. Research reviewed: 2026-10-24; evidence date: not recorded Unknowns: OpenAI has not mapped the current image model to a complete source inventory or documented its exact inherited base-model training recipe. ```json { "method": "training-rights-evidence-v1", "scope": "ChatGPT Images 2.5 feature, using OpenAI’s first-party image-generation route; training-data mapping is only to OpenAI’s broad system disclosures.", "components": { "provenance": { "score": 1, "reason": "The provider gives broad image-inclusive system data categories but no specific image model dataset inventory.", "sourceUrls": [ "https://help.openai.com/en/articles/20001044-training-data-summary-pursuant-to-california-civil-code-section-3111", "https://openai.com/policies/how-chatgpt-and-our-foundation-models-are-developed/" ] }, "permissions": { "score": 1, "reason": "OpenAI describes licensed partnerships alongside public sources, without model-specific permission coverage for this image route.", "sourceUrls": [ "https://help.openai.com/en/articles/20001044-training-data-summary-pursuant-to-california-civil-code-section-3111", "https://openai.com/policies/how-chatgpt-and-our-foundation-models-are-developed/" ] }, "creatorControls": { "score": 0, "reason": "The reviewed image and training disclosures do not establish creator remedies for the exact image-generation corpus.", "sourceUrls": [ "https://openai.com/index/introducing-chatgpt-images-2-5/", "https://help.openai.com/en/articles/20001044-training-data-summary-pursuant-to-california-civil-code-section-3111" ] } }, "routeDependency": null, "modelReference": null, "meaning": "Documented training-data rights safeguards. Low points may reflect missing evidence; they do not establish infringement." } ``` - [Introducing ChatGPT Images 2.5](https://openai.com/index/introducing-chatgpt-images-2-5/) (primary; checked 2026-10-24) - [Training Data Summary under California Civil Code Section 3111](https://help.openai.com/en/articles/20001044-training-data-summary-pursuant-to-california-civil-code-section-3111) (primary; checked 2026-10-24) - [How ChatGPT and our foundation models are developed](https://openai.com/policies/how-chatgpt-and-our-foundation-models-are-developed/) (primary; checked 2026-10-24) ## Your data & privacy Finding: 5/10 Assessment: https://ethicalstack.ai/products/chatgpt/#use-image-generation-privacy For consumer ChatGPT, prompts and images may be used to improve models depending on Data Controls; users can opt out for new conversations. Reason: This score applies to the consumer ChatGPT configuration. Content remains stored on OpenAI and service-provider systems, and authorized access can occur for stated support, safety, legal, or improvement purposes. Argument: This score applies to the consumer ChatGPT configuration. Content remains stored on OpenAI and service-provider systems, and authorized access can occur for stated support, safety, legal, or improvement purposes. Type: conditional; confidence: medium Scope (dimension): Selected configuration Research reviewed: not recorded; evidence date: not recorded Unknowns: This score applies to the consumer ChatGPT configuration. Content remains stored on OpenAI and service-provider systems, and authorized access can occur for stated support, safety, legal, or improvement purposes. - [Data Usage for Consumer Services FAQ](https://help.openai.com/en/articles/7039943-data-usage-for-consumer-services-faq) (primary; checked 2026-10-24) - [How ChatGPT learns about the world while protecting privacy](https://openai.com/index/how-chatgpt-protects-privacy/) (primary; checked 2026-10-24) ## Openness Finding: 2/10 Assessment: https://ethicalstack.ai/products/chatgpt/#use-image-generation-openness ChatGPT Images is documented as a feature of the hosted ChatGPT service. OpenAI publishes a system card, but the image model weights and implementation are not released for inspection or self-hosting. Reason: The product offers useful capability documentation while keeping the model and inference service proprietary. Users cannot inspect or move this model to their own infrastructure, so it fits the low openness range. Argument: The product offers useful capability documentation while keeping the model and inference service proprietary. Users cannot inspect or move this model to their own infrastructure, so it fits the low openness range. Type: evidence; confidence: medium Scope (dimension): Product Research reviewed: not recorded; evidence date: 2026-10-24 Unknowns: The product offers useful capability documentation while keeping the model and inference service proprietary. Users cannot inspect or move this model to their own infrastructure, so it fits the low openness range. - [ChatGPT Images 2.5 System Card](https://deploymentsafety.openai.com/chatgpt-images-2-5/safety-evaluations) (primary; checked 2026-10-24) - [OpenAI Service Terms](https://openai.com/en-GB/policies/service-terms/) (primary; checked 2026-10-24) ## Worker protections Finding: 4/10 Assessment: https://ethicalstack.ai/products/chatgpt/#use-image-generation-labour OpenAI’s supplier code sets pay, safety and complaint rules for data workers. Public evidence of coverage and outcomes is limited. Reason: 1 + 1 pay & conditions + 1 worker voice & remedy + 1 evidence & follow-through = 4/10. This grades documented worker safeguards, not a measured rate of mistreatment. Argument: The October 2026 Supplier Code applies to suppliers, staffing agencies, subcontractors and sub-tier suppliers, and explicitly covers annotation, moderation, evaluation and RLHF workers: it requires task-risk notice, breaks, exposure limits/support, task opt-out or escalation without retaliation, reasonable paid performance expectations, and supplier records that OpenAI may validate. A 2023 TIME investigation into the earlier Sama contract documented workers’ accounts of very low take-home pay and distressing work; Sama disputed key characterizations and said it offered extensive counseling, while OpenAI said it had understood workers could opt out and access support, and said vendors set pay and care. These historical claims and counterclaims do not verify the current code’s implementation, worker voice or outcomes. Type: argument; confidence: medium Scope (dimension): OpenAI's workforce and disclosed contractors. Coverage of individual product teams may be incomplete. Research reviewed: 2026-10-24; evidence date: 2021-2022 Unknowns: The current code does not publish supplier-level audit results, pay distributions, annotator coverage, worker survey results or independent remediation outcomes. The Sama project ended in 2022; no current vendor footprint is established. ```json { "method": "worker-safeguards-evidence-v1", "entity": "openai", "company": "OpenAI", "scope": "OpenAI's workforce and disclosed contractors. Coverage of individual product teams may be incomplete.", "components": { "conditions": { "score": 1, "reason": "The October 2026 supplier code sets specific conditions for human-data workers, but it is a written requirement rather than evidence of current implementation. The only cited implementation account is the historical GPT-4-era work; TIME’s 2022 Sama reporting documented worker concerns and OpenAI/Sama disputed or qualified key claims.", "sourceUrls": [ "https://openai.com/policies/supplier-code/", "https://cdn.openai.com/papers/gpt-4.pdf", "https://time.com/6247678/openai-chatgpt-kenya-workers/", "https://www.sama.com/blog/ethical-and-sustainable-sourcing" ] }, "voice": { "score": 1, "reason": "The current code requires channels to pause, decline or escalate unsafe tasks without retaliation, but no public evidence verifies a current feedback channel’s use or outcomes. Historical reporting records workers raising concerns and the vendor/customer responses, not current implementation.", "sourceUrls": [ "https://openai.com/policies/supplier-code/", "https://time.com/6247678/openai-chatgpt-kenya-workers/", "https://www.sama.com/blog/ethical-and-sustainable-sourcing" ] }, "assurance": { "score": 1, "reason": "The code requires suppliers to retain worker and contract records and cooperate with OpenAI or authorized third-party validation on reasonable request, but OpenAI publishes no worker-specific audit results or independent outcomes.", "sourceUrls": [ "https://openai.com/policies/supplier-code/" ] } }, "disclosureLimited": false, "meaning": "This grades documented worker safeguards, not a measured rate of mistreatment." } ``` - [OpenAI Supplier Code](https://openai.com/policies/supplier-code/) (primary; checked 2026-10-23) - [GPT-4 Technical Report](https://cdn.openai.com/papers/gpt-4.pdf) (primary; checked 2026-10-24) - [OpenAI Used Kenyan Workers on Less Than $2 Per Hour to Make ChatGPT Less Toxic](https://time.com/6247678/openai-chatgpt-kenya-workers/) (reporting; checked 2026-10-23) - [Sama and Content Moderation: What You Need to Know](https://www.sama.com/blog/ethical-and-sustainable-sourcing) (primary; checked 2026-10-24) ## Military ties Finding: 3/10 Assessment: https://ethicalstack.ai/products/chatgpt/#use-image-generation-military-contracts OpenAI describes a classified Department of War agreement with cloud-only deployment, named technical and personnel safeguards, and explicit restrictions, while lawful military use remains permitted. Reason: 1 + 1 use restrictions + 2 enforcement & oversight + 2 relationship disclosure = 6, capped at 3/10. Known defense ties lower this rating; safeguards can improve it within the relationship’s cap. Argument: The March 2026 agreement permits lawful purposes and classified deployment, so this is broad defense enablement; the public agreement does not identify actual targeting, weapons, or combat-intelligence deployments. It bars directing autonomous weapons and domestic surveillance of U.S. persons, including commercial personal data, and disallows high-stakes automated decisions where human approval is required. OpenAI says it retains the safety stack, deploys only through its cloud, uses classifiers and cleared personnel in the loop, and may terminate for breach. These are concrete company-described safeguards, but no independent enforcement outcomes are public, and the March amendment says NSA service would require a separate agreement. Type: argument; confidence: medium Scope (dimension): OpenAI's relationships and safeguards. A company relationship does not identify every product or release used. Research reviewed: 2026-10-24; evidence date: 2026-02-28 Unknowns: The public agreement does not list task orders, deployed model versions, users, operational workflows, audit outcomes, or whether guardrails have been independently tested in use. Contract permission and deployment of a specific model version are not established here. ```json { "method": "military-ties-priority-v2", "entity": "openai", "company": "OpenAI", "scope": "OpenAI's relationships and safeguards. A company relationship does not identify every product or release used.", "components": { "restrictions": { "score": 1, "reason": "The contract explicitly excludes autonomous weapon direction, domestic surveillance of U.S. persons, and specified unapproved high-stakes decisions, but lawful military activity remains authorized and not all harmful surveillance/weapon uses are categorically barred.", "sourceUrls": [ "https://openai.com/index/our-agreement-with-the-department-of-war/" ] }, "oversight": { "score": 2, "reason": "The agreement describes cloud-only deployment, OpenAI-controlled safety stack and classifiers, cleared staff in the loop, and termination rights; these are use-specific contractual and technical review mechanisms, with no independent outcome evidence.", "sourceUrls": [ "https://openai.com/index/our-agreement-with-the-department-of-war/" ] }, "transparency": { "score": 2, "reason": "OpenAI publishes the agreement’s scope, restrictions, architecture and NSA carve-out; it still does not disclose deployed mission/use-level details.", "sourceUrls": [ "https://openai.com/index/our-agreement-with-the-department-of-war/" ] } }, "relationship": "broad-defense", "relationshipReason": "A classified DoW contract authorizes broad lawful use, but the public evidence describes no actual targeting, strike, or combat-intelligence support; broad-defense cap applies.", "score": 3, "baseScore": 1, "safeguardPoints": 5, "safeguardAdjustment": 5, "rawScore": 6, "cap": 3, "disclosureLimited": false, "meaning": "Known military relationships take priority over missing safeguards. The rating does not establish the probability of an undisclosed contract." } ``` - [Our agreement with the Department of War](https://openai.com/so-DJ/index/our-agreement-with-the-department-of-war/) (primary; checked 2026-10-24) - [Our agreement with the Department of War — OpenAI](https://openai.com/index/our-agreement-with-the-department-of-war/) (primary; checked 2026-10-24) ## BDS targets Finding: Not listed Assessment: https://ethicalstack.ai/products/chatgpt/#use-image-generation-israel-palestine Categorical check of the BNC campaign guide and No Tech campaign page; absence from these named-target pages is not BDS approval. Reason: No named target for this company was found in the two checked BNC sources as of 2026-10-24. This is a narrow list check, not BDS approval, and it does not capture unlisted relationships or partners. Argument: No named target for this company was found in the two checked BNC sources as of 2026-10-24. This is a narrow list check, not BDS approval, and it does not capture unlisted relationships or partners. Type: conditional; confidence: medium Scope (dimension): Company record Research reviewed: not recorded; evidence date: not recorded Unknowns: No named target for this company was found in the two checked BNC sources as of 2026-10-24. This is a narrow list check, not BDS approval, and it does not capture unlisted relationships or partners. - [Guide to BDS Boycott & Pressure Corporate Priority Targeting](https://bdsmovement.net/Guide-to-BDS-Boycott) (primary; checked not recorded) - [No Tech for Oppression, Apartheid or Genocide](https://bdsmovement.net/no-tech-oppression-apartheid-or-genocide) (primary; checked not recorded) ## Reading the evidence - Assessments are provisional editorial judgments. Cite the profile and the underlying sources, and retain confidence, scope and unknowns. - Unknown, not-applicable and model-dependent are different states. Missing evidence does not establish good or bad conduct. - Company-level evidence does not establish use of a particular product or model. Account plan, model, host and route can change an assessment. - Energy scores describe documented practices and transparency, not measured request emissions. Zero component points can mean an evidence gap. - Published overall scores require at least four of six numerical dimensions. Preserve coverage and BDS adjustments; BDS itself is categorical. - No named BDS target found is limited to the reviewed campaign sources and is not approval. Personalized priorities do not change the published overall score. - Review and source-check dates describe the research, not this deployment. Fetch current primary sources when the decision requires current verification. Methodology: https://ethicalstack.ai/methodology/