Adobe Firefly

Adobe’s cloud creative-generation suite for image creation, editing and video; its creator-rights claims are narrower than universal contributor opt-out.

Score calculation

(5 + 7 + 7 + 2 + 7 + 4) ÷ 6 = 5.3

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

Full scoring method ↗

Setup details

Adobe Firefly: cloud access. Free tier and paid plans with generative credits; credit and subscription pricing vary by region. Ratings describe the stated product, company or configuration in each finding.

What fits your needs

Cloud Firefly web and Creative Cloud features; model, credit consumption and content terms vary by workflow and plan.

Tasks

Image generation, editing and video features connect to Adobe creative apps; workflows can include text/image inputs and generative editing.

See the evidence

Billing

Free access is limited; subscriptions use plan-specific generative credits, which renew monthly and do not roll over.

See the evidence

Generative credits FAQ ↗

Tradeoffs to expect

Route

Firefly web/app generation is cloud processed; there is no general local-only Firefly model route documented here.

See the evidence

Approach to generative AI with Adobe Firefly ↗

Training scope

Adobe says Firefly foundational models use licensed and public-domain material and are not trained on customer content; this is Adobe's stated model scope, not universal creator consent for all partner models.

See the evidence

How to switch

Get started

Use Firefly web or an eligible Creative Cloud app, sign in with an Adobe account, and check credit balances and whether the selected workflow uses Firefly or a partner model.

See the evidence

Practical questions

Age and sensitive content

Adobe GenAI guidelines prohibit explicit nudity, sexualized minors, impersonation, privacy-rights violations and harmful deception. Partner models have an 18+ minimum under Adobe’s product-specific terms.

See the evidence

Reference rights and likeness

Adobe guidelines prohibit using third-party personal information or references in violation of privacy, publicity or other rights; users must have rights for uploaded references.

See the evidence

Adobe Policy ↗

Output provenance

Adobe says qualifying Firefly exports automatically apply C2PA Content Credentials; users must not remove or alter credentials. Credentials indicate provenance metadata, not truth or permission.

See the evidence

Commercial use

Adobe labels its own Firefly models commercially safe; partner models shown in Adobe tools have separate terms, and any indemnity is limited to eligible workflows and offers.

See the evidence

Custom Models scope

Adobe says Custom Models learns from user-provided images and requires an explicit permission-to-train step. This is user-directed custom-model training, distinct from Adobe’s base-model training claim; partner models in Firefly have separate terms.

See the evidence

Energy practices

5/10↑

Adobe reports company-wide energy and renewable-electricity figures, while Firefly offers model, output-count, aspect-ratio, and resolution choices. No Firefly generation footprint is measured.

See the evidence

Adobe reports company-wide energy and renewable-electricity figures, while Firefly offers model, output-count, aspect-ratio, and resolution choices. No Firefly generation footprint is measured. The score describes documented measurement transparency and available controls, not environmental performance.

1 + 2 reporting + 0 electricity + 2 controls = 5/10. This grade covers documented practices and transparency; it does not measure the energy or emissions of a request.

Adobe’s CSR metrics cover company operations, not a Firefly or partner-model request. Firefly supports partner models and the serving route is not declared for a generation; annual renewable electricity doesn’t show hourly/location-matched power for this product.

Reporting2/3: Adobe’s 2024 CSR report publishes dated, scoped company energy and renewable-electricity metrics with reporting notes. These are company-wide rather than Firefly-specific, which caps this component at 2.
Electricity sourcing0/3: Firefly lets users select Adobe or partner models, and the reviewed docs do not declare the serving route or clean-power sourcing for the selected generation. Adobe’s annual corporate metrics cannot be allocated to a Firefly or partner-model request.
Resource controls2/3: Firefly supports model selection, aspect ratio, content type/composition, resolution in the current workspace, and one-versus-four image variants depending on model. These practical controls can bound output quantity/size, but Adobe reports no physical-resource reduction from using them.

Water disclosure

Water reporting has not been assessed for this choice.

Training data

7/10↑

Adobe’s public family policy describes licensed and public-domain sourcing, but its model-specific Image Model 5 summary adds OpenImages v7, privately licensed image/video data, separate private third-party image/video datasets outside the transactional-license category, and synthetic data from Firefly Image Model 5 and an unnamed internal model. The summary does not bound the rights basis or share of those latter categories.

See the evidence

The model-specific EU summary identifies a large public dataset and provides useful lineage, but also reports private image/video datasets outside its transactional-license category and synthetic data from an unnamed internal model. That does not prove unlawful use; it prevents treating the full family training data as predominantly and boundedly licensed on current evidence. Permission therefore scores 2; advance Stock licensing remains a narrow creator control.

1 + 2 source transparency + 2 permission + 2 creator controls = 7/10. This grades documented training-data safeguards, not a legal ruling.

The identities, rights basis and proportions of private third-party datasets outside the transactional-license category; the internal synthetic-data model and its lineage; coverage of this model-specific disclosure across later Firefly releases; and independent verification remain unavailable. Partner-model options in Adobe apps are outside scope.

Source transparency2/3: Adobe’s Image Model 5 summary names OpenImages v7 and Flickr for a limited CC0/public-domain search, and identifies licensed, other private and synthetic source categories; other third-party datasets remain unnamed.
Documented permission2/4: Adobe documents licensed private image/video data and a source-contract permission basis, but its Image Model 5 summary separately reports private third-party image/video datasets outside that transactional-license category and synthetic data from an unnamed internal model. Their rights basis and share are not bounded; this is an evidence gap, not proof of unlawful use.
Creator controls2/2: Adobe’s inspectable Stock Contributor Agreement grants advance license rights, and Adobe says the terms permit training; this supports consent for Stock material, not the other private/synthetic categories in Model 5.

Data-rights rubric ↗

Your data & privacy

7/10↑

Adobe says Firefly base models are not trained on customer content. Its Custom Models feature separately lets users upload 10–30 images and train a custom model from them; cloud processing and retention details vary by feature and account terms.

See the evidence

The no-customer-content-training commitment applies to Adobe’s base Firefly models, not to a custom model a user elects to train on uploaded images. This review did not establish end-to-end retention and access settings for every Firefly workflow.

Openness

2/10↑

Firefly is offered as a hosted generation service with paid plans and credits. Adobe publishes information about its approach and terms, but not model weights or an implementation that users can inspect and self-host.

See the evidence

The commercial service and some product practices are documented, but the system itself remains proprietary and cloud-operated. That limits inspection, modification and portability under the openness rubric.

Worker protections

7/10↑

Adobe reports global employee metrics and a formal human-rights and supplier system, including ethics reporting, supplier screening, code requirements and stated audit rights.

See the evidence

Adobe’s policy applies to employees, partners and suppliers: it prohibits forced/child labor and discrimination, commits to prompt investigation and non-retaliation, and offers managers, People Resources, Compliance and an anonymous hotline route. Its 2024 report states that 100% of significant suppliers were human-rights screened and all relevant suppliers were required to meet RBA standards; Adobe says it may audit partners. However, it reports no RBA validation audits required for turnkey suppliers and this is company-reported rather than independent worker-level verification.

1 + 2 pay & conditions + 2 worker voice & remedy + 2 evidence & follow-through = 7/10. This grades documented worker safeguards, not a measured rate of mistreatment.

Supplier checks are self-reported/company-directed and downstream annotation or creative labor is not individually mapped. Survey satisfaction, pay equity and zero reported fatality metrics are disclosed, but do not prove broad independent labor outcomes.

Pay & conditions2/3: Adobe supplies concrete employee and supplier standards, reports 100% significant-supplier screening and procurement spend screened or contractually covered, but lacks independent broad worker outcomes and says no turnkey supplier audits were required.
Worker voice & remedy2/3: Employees can report to managers, People Resources, Compliance or an anonymous hotline, with promised investigation and non-retaliation; equivalent supplier worker voice/outcome coverage is not demonstrated.
Evidence & follow-through2/3: Annual metrics include supplier screening/clauses, employee satisfaction, attrition, gender pay parity and contractor/employee fatalities; these are meaningful but self-reported, not independent worker-level verification.

Workers rubric ↗

Military ties

4/10Limited defense work↑

Adobe has a named DoD licensing vehicle for commercial creative/document software and Firefly content rules that prohibit violent extremist and graphic violence content; this is limited noncombat support, not evidence of Firefly operational military use.

See the evidence

Adobe and DoD describe a Joint Enterprise License Agreement making creative, publishing and document tools available across the Department. Separately, Firefly Guidelines ban graphic violence/gore, terrorism and violent extremism in hosted creative generation, allow account disabling, and describe automated/manual abuse review. They do not prohibit defense customers or establish Firefly use within the JELA. This is limited-defense support, not direct combat or intelligence work.

1 + 1 use restrictions + 1 enforcement & oversight + 1 relationship disclosure = 4/10. Known defense ties lower this rating; safeguards can improve it within the relationship’s cap.

The reviewed record doesn’t show all current contracts, subcontractors, deployment configurations, actual use by end users, or contract enforcement outcomes. For downloadable/open releases, company-hosted terms do not bind independent downstream deployments.

Use restrictions1/4: Firefly hosted guidelines bar graphic violence/gore and terrorism/extremism, but do not prohibit military or intelligence purposes generally; the JELA provides no model-specific use restriction.
Enforcement & oversight1/3: Adobe states it may disable accounts and uses automated/manual abuse review for Firefly, while the JELA reflects formal procurement terms; neither source establishes military-specific review or public enforcement outcomes.
Relationship disclosure1/2: The JELA and Firefly Guidelines are public and disclose broad creative/document use and prohibited content, but not Firefly-specific DoD deployment or operational outcomes.

Adobe and DoD describe a Joint Enterprise License Agreement making Adobe creative, publishing and document solutions available across the Department. The materials describe workforce creative/document productivity, not combat operations, targeting, weapons or intelligence systems. This is a limited-defense support relationship and is capped at 6 by the rubric. The relationship limits this grade to 6/10.

Military rubric ↗

BDS targets

Not listed↑

Categorical check of the BNC campaign guide and No Tech campaign page; absence from these named-target pages is not BDS approval.

See the evidence

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.

All sources