A high-profile controversy over artificial intelligence startups, with OpenAI, creator of ChatGPT, suing Chinese AI startup DeepSeek for pirating its model and developing a competing AI platform, has fueled a lot of controversy and debate over intellectual property, ethical AI development, and a growing competitive AI environment worldwide.
OpenAI claims that DeepSeek engaged in “distillation,” a training technique for AI in which information is extracted out of a larger, more complex model in an effort to expand a smaller one in terms of its size and intelligence. Distillation is a common practice in AI development, but OpenAI claims that DeepSeek’s actions violate its terms of service in using OpenAI’s copyrighted models in developing a competing AI platform.
The accusations first arose when Microsoft security engineers detected suspicious file transfers between suspected OpenAI developer accounts for DeepSeek. OpenAI, in whose board of investors Microsoft holds a significant seat, alerted OpenAI to suspected unauthorized harvesting, and an investigation took place. OpenAI then disabled suspected accounts and is cooperating with the U.S. government in an attempt to prevent future misuse of its technology. That a case involves Microsoft shows its gravity and its potential for broader ramifications in the technology community.
DeepSeek hasn’t yet publicly commented on OpenAI’s claims, but its startup recently unveiled its AI model, DeepSeek-R1, which took center stage for its performance and price tag. According to its claims, DeepSeek developed R1 for a mere $5.6 million—an incredibly low budget when compared with OpenAI’s alleged $100 million+ outlay in developing GPT-4. That budget chasm fueled speculations that DeepSeek must have leveraged OpenAI’s intellectual property in developing such quick breakthroughs. Despite controversy, DeepSeek’s R1 received praise for its efficiency, with claims of its performance at par with best AI models, including OpenAI’s ChatGPT and Meta’s Llama 2.
The claims have surprised both technology and AI sectors. Initially, its worth tumbled in a free fall, cutting down a significant portion of its worth in terms of value in the marketplace, but then recovered partially later during the day with investors’ faith. That uncertainty mirrors both the AI environment’s interrelated character and its susceptibility to such scandals.
The case has even fueled a broader conversation regarding AI intellectual property rights. There have been claims in the community that numerous AI startups, including U.S.-based ones, utilize information derived from successful ones in an effort to make them even successful ones. Others have countered that OpenAI’s concerns have basis, particularly in consideration of AI competition and its accompanying tension in terms of geopolitics.
OpenAI took strong measures in response to accusations, including closing suspected accounts and enhancing security controls. OpenAI is working closely with the U.S. government in a bid to prevent foreign entities from developing U.S.-sophisticated AI clones in a non-consensual manner. OpenAI emphasized that Chinese AI entities have been actively working towards creating clones of U.S.-based AI, and national security is at stake. There have been claims that even the U.S. Navy prohibited its employees from using DeepSeek’s AI tools for fear of misuse of information at the hands of the Chinese government.
The controversy has rekindled discussion about ethics in AI use of information, as well. As OpenAI condemns DeepSeek for its supposed actions, it is defending similar claims in court, too. Media entities, including Canadian media companies and The New York Times, have sued OpenAI for taking copyrighted work and utilizing it to train AI algorithms in a manner not endorsed by them. That raises a larger question: Where can one draw a boundary in AI development? If OpenAI and similar entities utilize publicly available information to build and train AI algorithms, can they claim a prerogative to prohibit everyone else from following in its footsteps? The case brings out the ongoing tension between innovation, intellectual property, and ethics in AI training.