"Can we use copy generated by ChatGPT directly on our website and ads?" "If we use images generated with Gemini in corporate brochures or social posts, will that cause legal issues later?"—these are currently the most common concerns voiced by SMBs adopting generative AI for marketing. Teams understand its convenience. Yet ambiguity remains over who ultimately owns the text and images, and whether using them commercially without permission could prompt legal pushback. Consequently, work grinds to a halt.
To cut to the conclusion, this question cannot be answered with a blanket yes or no. Copyright status of the generated content, terms of service of the tools used, and whether outputs resemble existing protected works represent three distinct layers, each governed by different criteria. This article organizes these three layers based on Japanese official guidelines and major platform terms of service so non-legal professionals can independently evaluate key checkpoints. Please note that this article does not constitute legal advice; always consult qualified legal counsel for final decisions regarding critical applications.
Layer 1: Does AI-generated content hold copyright in the first place?
The first point to understand is whether AI-created materials possess copyright. Under Japanese copyright law, a copyrighted work is defined as a "creative expression of thoughts or feelings," protecting human creations. In the "Viewpoint on AI and Copyright" (2024) compiled by the Agency for Cultural Affairs' Subdivision on Copyright, the guiding principle states that when AI is used as a tool, outputs may qualify as copyrighted works if the user is recognized as having made a "creative contribution," whereas outputs autonomously produced by AI without creative input are not recognized as copyrighted works. The determination is made on a case-by-case basis, evaluating prompt substance and the extent of iterative effort.
A common misunderstanding among practitioners is assuming that "no copyright means it cannot be used." In fact, the opposite holds: the difficulty of establishing copyright over AI-generated outputs means that while you can use them, competitors may also be entirely free to use similar outputs. You might invest significant effort to produce a key visual only to find you hold no exclusive rights, leaving you unable to prevent a competitor from using an almost identical image. The more central a visual is to your brand identity, the more conscious you must be of this lack of protection. However, standards in this domain continue to evolve across jurisdictions and accumulating precedents; anchoring your criteria in official frameworks like those from the Agency for Cultural Affairs remains the safest approach.
Layer 2: Commercial use permissions are determined by terms of service
Distinct from copyright ownership, whether you can use a service's generated content commercially is governed by its terms of service. These conditions vary across providers and often differ between free tiers, paid plans, and API offerings within the same service.
For example, OpenAI's terms state that users retain ownership of their inputs, and OpenAI assigns to the user any rights it holds in the generated outputs, permitting commercial use. Google's Generative AI Additional Terms of Service likewise state that Google claims no ownership over original content generated by its services. Meanwhile, free tiers frequently impose restrictions on usage volume or data handling, or assume enterprise-tier plans when commercial use is intended. Detailed conditions vary by plan.
Crucially, you should verify the relevant clauses directly in the current terms of your active subscription, rather than relying on secondary articles or hearsay. Terms are updated periodically; usage permitted six months ago may not carry the same terms today. Rules established here will fail unless integrated with day-to-day operations specifying who uses which tools and how. Approaches to defining approved tools and plans at an organization level are outlined in practical establishment of internal AI usage rules and guidelines, while operational pitfalls including shadow AI are addressed in how to create generative AI usage guidelines. A practical approach is embedding this check into your internal policy by recording who reviewed the terms of each adopted tool and when.
Layer 3: Infringement risks originating from training data
The third layer is the most frequently overlooked and hazardous in practice: the risk that generative AI outputs closely resemble existing copyrighted works, characters, logos, or trademarks. Even if commercial use is allowed by terms of service and no third party holds copyright over the AI output, an output that coincidentally mirrors a real-world creation could constitute copyright or trademark infringement if used publicly. One cannot dismiss liability by claiming "the AI generated it"; liability rests squarely with the entity that uses and publishes the material.
Under Japanese law, copyright infringement is broadly evaluated based on reliance (access to the pre-existing work) and similarity of expression. In the Agency for Cultural Affairs' "Checklist & Guidance on AI and Copyright" released in 2024, recommended measures to prevent infringement include checking similarity between generated outputs and existing works, as well as avoiding usage patterns that suggest reliance. In particular, prompting AI to generate content in the style of specific characters, artists, or brand names creates significant resemblance risk and is dangerous for marketing collateral.
| Layer | What to evaluate | Core focus |
|---|---|---|
| Copyright in outputs | Can your company claim exclusivity? Can others use it too? | Official guidance from the Agency for Cultural Affairs / Case-by-case assessment |
| Commercial use permission | Does the specific service allow commercial utilization? | Terms of service of each service (by plan) |
| Infringement risk | Does it resemble existing creative works or trademarks? | Similarity and reliance verification |
Practical checklist before publishing
Based on these three layers, concrete checks emerge for practitioners. First, review the relevant clauses in your tool and plan's terms of service. Confirm commercial permissions and output ownership in primary source text rather than hearsay, recording the date of review. Second, perform similarity checks. Before publishing text or images intended for high-stakes applications, run reverse image or trademark searches to ensure they do not mirror existing works, particularly regarding logos, characters, and well-known slogans. Third, keep generation audit trails. Document which tool generated the asset, using which prompts, and on what date. This provides documentation to show content was not intentionally plagiarized if disputes arise. Principles of watermarking to trace AI content provenance are also explored in AI content provenance and digital watermarking.
Fourth, seek legal review for high-impact assets. For assets with significant business exposure and high switching costs—such as corporate brand logos, flagship packaging, or major ad campaigns—avoid relying solely on operational judgment and consult qualified legal counsel. Conversely, minor collateral like internal memos or working drafts does not require formal legal vetting. Calibrating review depth to asset criticality is the realistic, sustainable approach.
What to do next
The immediate actions following this article can be distilled to two steps. First, open and read the commercial use and ownership sections of the terms of service for the generative AI tools your company currently uses, specific to your subscription tier. Second, inventory any AI-generated assets already deployed in marketing and run similarity checks on high-impact items (logos, advertisements, packaging). These two actions reveal the majority of latent pitfalls. From there, you can establish operational models, such as deciding whether to generate promotional images in-house or outsource them, at your company's own pace by referencing in-housing vs. outsourcing decisions for AI-generated marketing imagery.
GleamHub specializes in creative production for websites and advertising alongside ongoing advisory for corporate AI adoption. If you want to use generative AI for marketing but have concerns regarding legal rights, or want to establish secure internal AI workflows, feel free to reach out through our development, AI, and automation consultation desk. We will help you identify key criteria for reviewing terms of service and design secure production pipelines. Please note that we recommend consulting legal counsel for definitive legal assessments regarding rights and ownership.
Sources
- About AI and Copyright | Agency for Cultural Affairs
- Viewpoint on AI and Copyright (March 15, 2024, Legal System Subcommittee, Copyright Division, Council for Cultural Affairs) | Agency for Cultural Affairs (PDF)
- Checklist & Guidance on AI and Copyright (July 31, 2024) | Agency for Cultural Affairs (PDF)
- Terms of Use｜OpenAI
- Generative AI Additional Terms of Service | Google
