In recent weeks, the AI community has been abuzz with controversy—and healthy debate—over claims that Chinese competitors are “stealing” OpenAI’s work to rapidly advance their own models. As discussions swirl on intellectual property rights, model replication, and ethical data use, it’s worth taking a step back to assess both the technical and ethical sides of the issue. This post explores what’s really happening, why it matters for innovation, and what it means for the future of AI development.

Disclaimer: While I strive for balance in this analysis, I acknowledge that inherent biases may exist in my perspective. Readers are encouraged to approach these complex issues with critical thinking and draw their own conclusions.

The Heart of the Debate: Data, Distillation, and Disclosure

At the core, OpenAI has raised concerns that rivals—DeepSeek among them—might be leveraging its work to build competitive AI systems. OpenAI argues that these competitors are “distilling” its models, meaning they extract knowledge (or “distill” it) from OpenAI’s output to train their own models. But when we break down what this means in practice, things get more nuanced.

On the technical side, strict knowledge distillation typically requires access to rich, internal signals from a large model—detailed token probabilities, intermediate activations, and other “thinking” tokens that reveal how the model arrived at its answer (see my previous post for details). However, in most real-world settings, including the scenario with DeepSeek, competitors only have access to the final output provided via an API. This means that instead of full-blown internal distillation, what is likely happening is a form of supervised fine-tuning (SFT) on those final “hard” outputs. And that isn’t exactly a novel or ethically questionable technique; it’s a standard practice in AI research and development.

Technical Realities: What Can (and Can’t) Be Copied

For anyone steeped in the world of AI, it’s clear that replicating a state-of-the-art model like OpenAI’s requires more than just mimicking its final responses. OpenAI’s newer models use sophisticated inference-time scaling techniques that deliberately hide internal “reasoning” steps, making it virtually impossible to reverse-engineer the underlying processes without insider access. DeepSeek’s approach, by contrast, appears to rely on training using the public outputs of ChatGPT-like systems—a method that aligns with common industry practice.

In essence, if DeepSeek is indeed using only the outputs available through the API, then what they’re doing amounts to supervised fine-tuning. That is, they take a model’s final predictions and use those as labels to train their own system. While OpenAI’s terms of service may have strict rules about building competing products using their API outputs, the technical reality remains: using hard outputs for further training is a well-accepted technique that has long been part of the research playbook.

The ethical debate, however, is not solely about technical replication. There’s a broader issue at play regarding intellectual property rights and the consent of content creators. OpenAI, like many other companies, relies heavily on publicly available data to train its models. While this is legally defended under doctrines such as fair use, it places the industry in a murky ethical area. Critics contend that even if the data are public, using them without explicit permission or compensation may undermine the rights of the original creators.

Interestingly, OpenAI itself has not been immune to similar criticisms. Over the years, the company has faced numerous lawsuits and ethical challenges regarding its use of copyrighted material—from news articles to books. This duality raises an important question: How do we balance the need for large, diverse datasets that drive AI innovation with the imperative to respect the creative rights of individuals and organizations?

Many argue that the current industry practices, including both OpenAI’s and DeepSeek’s, highlight the need for clearer licensing frameworks and more transparent data usage policies. Rather than singling out one competitor for using a standard training method, it may be more productive to address the broader challenge of establishing ethical and legal norms in an era when data—and how we use it—is king.

Looking Ahead: Innovation, Accountability, and Transparency

So, what does this mean for the future of AI? On one hand, the technical methods that underpin our models—whether through full knowledge distillation or supervised fine-tuning—will continue to evolve. The industry is actively developing better techniques to both safeguard internal model processes and to train new models efficiently. On the other hand, the ethical and legal landscape is still catching up. As more stakeholders—from artists to publishers—raise their voices, we can expect increased calls for accountability and clearer standards on data usage.

What remains crucial is a balanced perspective. While technical safeguards (like hiding internal “thinking” tokens) are important, they do not negate the ethical imperatives to ensure that the original creators of content are respected and, ideally, compensated. A collaborative effort toward transparent data policies and licensing arrangements might be the best path forward for all parties involved.

Final Thoughts

The debate between OpenAI and its competitors such as DeepSeek isn’t simply a story of one company accusing another of copying—it’s a microcosm of larger issues facing the AI industry today. As we continue to push the boundaries of what artificial intelligence can do, we must also be mindful of the ethical, legal, and technical frameworks that underpin these innovations. By engaging in balanced discussions and working towards more transparent practices, we can ensure that the AI revolution benefits everyone while respecting the rights of content creators.

In the end, the conversation is far from over. It’s an ongoing dialogue that challenges us to rethink not only how we build AI systems, but also how we respect and value the human creativity that fuels them.