Chinese startup DeepSeek made waves this month by launching an AI model that appeared to rival those from Silicon Valley, utilizing fewer resources and creative workarounds to U. S. chip restrictions. Initially hailed as an engineering triumph, the narrative has shifted with OpenAI alleging that DeepSeek may have improperly used its proprietary data for training its AI. OpenAI is investigating whether DeepSeek mimicked its model outputs in a process called distillation, which, although common in AI development, violates OpenAI’s terms of service. Support for OpenAI's claims has surfaced from U. S. officials. Commerce secretary nominee Howard Lutnick accused DeepSeek of misusing U. S. technology, while David Sacks from the White House acknowledged possible intellectual property theft.

This controversy highlights a significant question about the control companies should have over their AI when derived from data sourced from others. The situation is particularly ironic for OpenAI, which faces lawsuits for allegedly using copyrighted material for training its AI and is criticized for its own data practices. Despite DeepSeek’s acknowledgment of using some open-source models, OpenAI claims that it improperly utilized closed-source models, though evidence of wrongdoing remains unpresented. Proving improper distillation could be challenging for OpenAI, as any claims might hinge on disclosing its own model training details. Legal experts are skeptical about the viability of OpenAI's case against DeepSeek, citing the difficulties in enforcing terms of service across borders. The financial implications of DeepSeek's advancements have caused fluctuations in U. S. tech stocks, prompting conversations about potential government actions, such as tariffs, to curb technology theft. Contradicting these concerns, President Trump praised DeepSeek's low-cost model as a positive development for AI, suggesting it might encourage competitiveness in the industry. Overall, if DeepSeek’s model proves sustainable, it could reshape the market dynamics, opening up opportunities for smaller teams to innovate in the generative AI field.