Generative AI is everywhere. It’s being embedded into creative workflows, marketing platforms, content tools, and automation systems at a breakneck pace. Tech vendors and service providers are rolling out AI-powered features wrapped in sleek interfaces, promising faster, cheaper, and more efficient ways to create. But beneath the polished sales pitches lies a growing risk that many businesses are sleepwalking into: the AI models powering these tools are often built on scraped, unlicensed, and legally questionable data. And when intellectual property (IP) is compromised at this scale, it doesn’t just affect creators, it puts businesses in legal and financial jeopardy. And that risk? It’s not just theoretical - it could land brands, agencies, and entire industries in serious trouble.
The Problem Hiding in Third-Party AI Models Right now, most AI tools being integrated into creative workflows -whether in advertising, publishing, design, or media - are not built in-house by the companies selling them. Instead, many vendors are taking open-source or third-party AI models, wrapping them in a user-friendly interface, and offering them as premium services. The problem? Many of these AI models have been trained on vast amounts of copyrighted content without consent.
The AI generating images? It might have been trained on millions of copyrighted photos, illustrations, and artworks without the original artists’ permission.
The AI writing copy? It could be pulling from thousands of books, articles, and scripts—none of which were licensed for use.
The AI producing music? It may have been trained to mimic styles from musicians who never agreed to have their work repurposed.
If a brand uses these tools without understanding what’s under the hood, they’re exposed to legal, ethical, and reputational risks. Worse still, anything generated by these AI models cannot be classified as original IP. Since the output is a direct result of existing copyrighted materials, it lacks the originality and independent creation required to be legally recognised as intellectual property. The Legal and Ethical Time Bomb AI companies are already facing lawsuits from creators whose work has been used without consent. Getty Images is suing Stability AI for allegedly scraping millions of its images. Writers and artists are challenging AI platforms for replicating their work without credit or payment. For brands and agencies, the danger is clear: if the AI tool you’re using is built on unlicensed data, you’re benefiting from copyrighted material without permission. And while AI vendors might argue that “transformative use” protects them, that’s not a legal shield for businesses that rely on these tools. The rules around AI-generated content are shifting fast, and brands using unchecked models today could find themselves dealing with serious consequences tomorrow. Beyond the legal risks, there’s a commercial risk that businesses haven’t yet fully grasped: if AI-generated content can’t be classified as IP, then brands lose ownership over the assets they create. A marketing campaign, a brand identity, a product description - if AI generates it, who actually owns it? Without proper protections, businesses risk investing in creative assets they have no legal rights to control. Technology Partners Need to Be Transparent About Their AI Choices Businesses need to start asking tougher questions about the AI models powering their tools:
What training data was used? If a vendor can’t answer this clearly, that’s a red flag.
Is the AI model open-source, proprietary, or licensed? If it’s a third-party model with no clear ownership, assume there’s risk.
Is there indemnification against copyright claims? If the vendor isn’t willing to back their product legally, why should your business take the fall?
Does the vendor offer first-party AI solutions? If they’re just repackaging generic AI models, you need to reconsider the trust you’re placing in them.
The brands that protect themselves will be the ones that demand AI transparency from their technology partners, not just glossy demos and productivity promises. The Safer Path: First-Party and Licensed AI Models The good news? AI can be built ethically and legally. The AI companies with real long-term strategies, the ones businesses should be working with are building models using first-party data, licensed datasets, and transparent methodologies. These models might cost more to develop, but they offer something no scraped-together AI service ever will: legal and ethical certainty. And here’s the thing, AI companies aren’t short on cash. OpenAI has raised over $11 billion, Anthropic $7 billion, and Stability AI is valued at over $1 billion. If these companies were serious about ethical AI, they’d be licensing data properly. Many of them simply don’t want to. Businesses should take note: if an AI model is free to train on everything, it’s probably free from accountability too. The Bottom Line: Protect Your Brand Before It’s Too Late The AI revolution is happening. It’s not a question of whether businesses will use AI, but how responsibly they do it. Brands, agencies, and media companies must stop blindly adopting generative AI tools without understanding the risks of the models that power them. If technology partners can’t provide clear answers about their AI models, businesses need to rethink their reliance on those tools. Otherwise, they risk being on the wrong side of the next major AI lawsuit, regulatory shift, or reputational fallout. Because in the end, AI isn’t just a tool - it’s a liability if you don’t know where it came from. And if it compromises IP, it compromises everything.
