Why has patent infringement not been in the news about DeepSeek?

DeepSeek hit the news recently with its large language model (LLM), which has striking similarities to ChatGPT, the well-known Artificial Intelligence (AI) model developed by OpenAI. There has been a lot of discussion and speculation about how DeepSeek has been able to produce similar results so quickly and on a significantly lower budget. This has also raised questions about what this means for the perceived US dominance when it comes to development in AI.

Despite the apparent similarities between the outputs generated by the DeepSeek model and ChatGPT, there have (so far) been no allegations of patent infringement against DeepSeek. There are perhaps two reasons for this. First, protecting high-level AI concepts is inherently challenging. Second, OpenAI’s relatively limited patent filing strategy leaves it with fewer legal options. We will explore these factors below.

The challenges of patenting high-level AI concepts

One reason why no patent infringement claims have emerged could be due to the difficulty in securing broad protection for high-level AI concepts. Companies like Google and Facebook have filed many patent applications, but have been unable to obtain patent protection for fundamental ideas such as a search engine or a social media platform. Broad and high-level concepts such as these are generally difficult to protect with patents. The history of technology giants, such as Yahoo giving way to Google, and MySpace being overtaken by Facebook, illustrates how first-mover advantage does not guarantee long-term dominance in a broad field.

This same principle is now repeating in the LLM space, where a first mover has not been able to prevent competition by others. Despite OpenAI’s pioneering role in advancing key LLM concepts, high-level concepts are challenging to protect through patents. As a result, competitors like DeepSeek have been able to enter the space apparently without significant freedom to operate issues.

The limitations of OpenAI’s patent strategy

On the second reason, it is notable that OpenAI has only filed around 14 patent applications, primarily in the US, with a few international (PCT) applications still pending. Given the size and reputation of ChatGPT, this number seems surprisingly low. It seems that most of OpenAI’s patent applications are directed towards specific use cases, suggesting that OpenAI may lack broad patents that could be used to challenge DeepSeek and other competitors.

One possible explanation for this limited patent strategy is OpenAI’s assumption that its main competitors would be US-based tech giants. Developing and operating LLMs requires vast resources, including high-performance semiconductor chips and data centres. Historically, only tech companies such as Google and Microsoft had the capacity to sustain such large-scale AI systems. It seems possible that this contributed to the popular assumption that China, and other countries, would struggle to compete.

DeepSeek has upended this assumption by developing its AI model that operates with significantly fewer chips and on a lower budget. Given this, OpenAI may need to reconsider its patent strategy, since competitors like DeepSeek have demonstrated that AI development is possible with more limited resources.

The future of AI protection

The case of DeepSeek versus OpenAI highlights the rapid nature of AI development. Looking ahead, the AI industry may continue to see increased competition, with new models emerging from various parts of the world. The perceived dominance of US-based AI firms is being tested, and companies like DeepSeek demonstrate that AI development is no longer confined to Silicon Valley.

It is hard to say whether OpenAI’s limited patent portfolio is based on a strategic decision to limit its patent filings. But it is clear that the lack of broad patents has enabled competition, raising questions about whether a different approach would have enabled OpenAI to achieve greater dominance in the industry.

For AI companies, considering these factors when developing intellectual property strategies will prove to be important. As AI technology continues to evolve, finding the right balance between open innovation and legal protection will be a hot topic of conversation.