Entrepreneurs and businesses need to share ideas, designs and expertise to market their businesses, educate customers, and demonstrate what makes them different. But once something is publicly available, controlling how it is subsequently accessed or used becomes much more difficult. AI adds a new dimension to that familiar IP challenge, and another reason to think carefully about what you publish and how you protect it.
Organizations publish a staggering (and often under-acknowledged) amount of proprietary and original content. Product imagery, technical illustrations, research, training materials, marketing assets and educational resources all play an important role in communicating expertise and supporting growth.
At the same time, publishing content online means giving up a certain degree of control over where that content may ultimately appear or how it may be used.
Generative AI has added new considerations to the equation. Publicly available material may be collected as part of datasets used to train AI systems, raising questions for organizations about how they protect valuable content while continuing to make useful information accessible to customers, partners and other audiences.
Copyright remains an important piece of legal protection, but it's only one part of the broader IP picture. A new generation of tools such as Glaze and Nightshade show how technical safeguards are evolving alongside AI and how they could complement more traditional approaches to intellectual property protection.
Many entrepreneurs and mid-sized organizations invest significant time and resources creating proprietary content, often holding a significant amount of intellectual property in forms that may not be immediately thought of as “IP”:
Depending on the asset and circumstances, these materials may be protected through copyright or other forms of intellectual property.
The more practical challenge is that owning rights to an asset does not necessarily give an organization complete control over how publicly accessible material is collected or used by others.
Organizations that believe their rights have been infringed may face an enforcement process that is difficult, expensive, drawn out, and that often occurs only after the content has already been accessed and used.
Protecting IP in an AI-enabled environment increasingly requires thinking about how an asset will be exposed and used, not simply whether you legally own it. As a result, many creators and organizations are looking for preventative measures rather than relying solely on legal remedies.
Developed by researchers at the University of Chicago, Glaze and Nightshade are tools designed to interfere with how AI systems learn from content. Although they were created primarily with artists in mind, they illustrate a broader approach that may be relevant to organizations thinking about how publicly available intellectual property could be used in AI training.
| Glaze: The Defensive Shield | Nightshade: The Active Deterrent |
|
Glaze modifies digital images in ways that are largely invisible to the human eye while causing AI systems to misinterpret stylistic characteristics. The goal is to make it more difficult for generative AI models to accurately learn and reproduce a creator's distinctive style. Think of Glaze as a form of digital camouflage. A human viewer still sees the original image, but an AI model processing the image "sees" a different style signature. If the image is later included in a training dataset, the model learns inaccurate information about the underlying artistic style. |
Nightshade uses a different technique and takes protection one step further. Rather than merely hiding stylistic information, it intentionally introduces data that can cause AI models to learn incorrect associations, an approach that researchers describe this as a form of "data poisoning." For example, an AI model trained on sufficient Nightshade-protected content may begin making incorrect connections between concepts, reducing the value of scraped datasets. The objective is not to damage AI models entirely but to increase the cost and risk of training on content that has been collected without permission. |
Although Glaze and Nightshade are relatively specific in their application, their broader value may be in prompting organizations to think more carefully about what they make publicly available in the first place..
A technical diagram, for example, may help explain a product to customers while also revealing information about how that product works. Training content might help customers use a service while also capturing valuable organizational know-how. Original product imagery or designs may support marketing while representing creative assets with value of their own.
The main lesson here should be to ask the question: “How much thought goes into what we are publishing before an asset goes online?" This can include considerations such as:
1. Protecting Proprietary Visual Assets These assets may have value beyond their immediate marketing purpose. Applying AI-resistant techniques before publishing these assets can help reduce their usefulness in future training datasets.
2. Preserving Competitive Advantage For startups and technology companies in particular, proprietary content can often represent years of accumulated expertise.
If AI systems can easily absorb and reproduce that expertise, the competitive value of those assets may erode. Defensive technologies can provide an additional layer of protection alongside patents, copyrights, trade secrets, and contractual controls.
3. Supporting a Consent-Based AI Ecosystem Many creators are not opposed to AI itself. Rather, they object to having their work used without permission or compensation.
Tools like Glaze and Nightshade shift some power back to content owners by creating economic incentives for AI developers to license content legitimately rather than scrape it indiscriminately.
Tools like Glaze and Nightshade are not a complete copyright solution.
They Do Not Prevent Copying
Someone can still download, view, share, or manually copy the content. Glaze and Nightshade do not function like traditional digital rights management (DRM).
They Primarily Protect Images
The technology was designed for visual content. Most business documents, reports, presentations, white papers, and text-based materials cannot currently be protected using the same techniques.
They Are Not a Legal Substitute
Copyright registration, clear licensing terms, confidentiality agreements, trade secret practices, and patent protection remain essential components of an IP strategy. These tools should be viewed as complementary safeguards rather than replacements.
Before publishing a potentially valuable asset, teams may want to consider:
No single measure is sufficient, but together they create a stronger defensive posture.
Valuable IP is often created across multiple functions rather than sitting neatly within a single department. Marketing may own the publishing process while product, technical or leadership teams understand the underlying value of the information being shared.
Creating clearer internal practices around those decisions can help prevent valuable IP from being disclosed without the right people having considered the implications.
Glaze and Nightshade are still specialized tools, and their usefulness will depend on the type of content an organization creates and the risks it is trying to manage. But they represent an important shift in how creators think about intellectual property in the age of AI.
Rather than waiting for courts and regulators to resolve complex copyright questions, these tools offer a practical, technology-based response that helps creators retain greater control over how their work is used.
For organizations concerned about their content being used in AI training, there is unlikely to be a single tool or legal mechanism that resolves every concern. Protecting intellectual property increasingly requires both legal and technical strategies.
A more practical starting point is understanding which assets matter, how they are currently being shared and whether the protections around them reflect their value.
As AI creates new ways for publicly available information to be collected and reused, those questions are becoming part of the normal conversation around managing intellectual property.
In other words, putting an AI-focused safeguard on an image isn't an IP strategy. It's one potential tool within one.
AI didn't create the challenge of protecting valuable content once it's made public. But it gives organizations another reason to pay attention to it.
Not sure whether your current approach to IP reflects how your organization creates and shares valuable content? Talk to Stratford's IP team about building a practical IP strategy around the assets that matter to your business.
| Natalie Giroux is the founder and former President of Stratford Intellectual Property. Retiring from that role in 2025, Natalie continues to support Stratford as an executive advisor. With deep expertise in strategic IP management and a business-first approach, Natalie has supported over 100 companies in aligning their IP portfolios with growth objectives. She has been internationally recognized multiple times as a leading IP strategist, including being named to the IAM Strategy 300 list. She is passionate about maximizing the value of innovation. |