Executive Summary:
A study by AI detection firm Copyleaks reveals that 74.2% of the text generated by DeepSeek shows a significant stylistic similarity to outputs from OpenAI’s ChatGPT, suggesting that DeepSeek may have been trained on ChatGPT’s data. This finding raises potential concerns about intellectual property rights, AI regulations, and the future of AI development. The research, shared exclusively with Forbes before its publication, highlights the implications of AI models potentially deriving from existing technologies.
Key Points:
- Copyleaks’ study shows a 74.2% stylistic similarity between DeepSeek and ChatGPT.
- This similarity was not observed with other AI models such as Claude, Gemini, and Llama.
- The findings prompt questions about possible training practices of DeepSeek on OpenAI’s outputs.
- Significant implications for intellectual property rights and AI regulations are anticipated.
References:
- The study will be published on Cornell’s arXiv.org repository.
- Article link: DeepSeek’s AI Style Matches ChatGPT’s 74 Percent Of The Time—New Study
12ft.io Link: https://12ft.io/https://www.forbes.com/sites/torconstantino/2025/03/03/deepseeks-ai-style-matches-chatgpts-74-percent-of-the-time-new-study/
Archive.org Link: https://web.archive.org/web/https://www.forbes.com/sites/torconstantino/2025/03/03/deepseeks-ai-style-matches-chatgpts-74-percent-of-the-time-new-study/
Original Link: https://www.forbes.com/sites/torconstantino/2025/03/03/deepseeks-ai-style-matches-chatgpts-74-percent-of-the-time-new-study/
User Message: DeepSeek’s AI Style Matches ChatGPT’s 74 Percent Of The Time—New Study
for more on see the post on bypassing methods