OpenAI suspects that China's DeepSeek AI models, known for their low cost, were developed using OpenAI data.
President Donald Trump called DeepSeek a “wake-up call” for U.S. tech firms this week, as Nvidia lost nearly $600 billion in market value.
DeepSeek’s rise triggered a sharp decline in AI-focused company stocks. Nvidia, a leading GPU provider for AI, saw its shares plummet 16.86%, marking Wall Street’s largest single-day loss.
Microsoft, Meta Platforms, and Alphabet, Google’s parent, dropped between 2.1% and 4.2%, while Dell Technologies, an AI server manufacturer, fell 8.7%.
DeepSeek touts its R1 model as a cost-effective alternative to Western AI like ChatGPT. Built on the open-source DeepSeek-V3, it reportedly needs less computing power and was trained for approximately $6 million.
Though some question this claim, DeepSeek has raised concerns about the massive AI investments by U.S. tech giants, unsettling investors. The model’s popularity surged, topping the U.S. free app download chart amid growing buzz about its performance.
Bloomberg reported that OpenAI and Microsoft are investigating whether DeepSeek used OpenAI’s API to incorporate OpenAI’s AI models into its own. “We know PRC-based companies and others are persistently trying to extract leading U.S. AI models,” OpenAI told Bloomberg.
Distillation, a method where developers train AI by pulling data from larger models, violates OpenAI’s terms of service.
“As a top AI developer, we take steps to safeguard our IP, carefully selecting which advanced features to include in released models. Moving forward, we believe close collaboration with the U.S. government is vital to protect our most advanced models from adversaries and competitors,” OpenAI stated.
David Sacks, President Trump’s AI advisor, told Fox News: “There’s strong evidence DeepSeek distilled knowledge from OpenAI models, and OpenAI is not pleased. Expect leading AI firms to take action against distillation in the coming months.”
Critics noted the irony, as OpenAI has faced accusations of using internet data to build ChatGPT. Tech writer Ed Zitron tweeted: “I can’t stop laughing. OpenAI, built on scraping the entire internet, is upset because DeepSeek might have used ChatGPT outputs for training. They’re acting like victims—what hypocrisy.”
In January 2024, OpenAI stated it was “impossible” to develop tools like ChatGPT without copyrighted material.
In a submission to the UK’s House of Lords communications and digital select committee, OpenAI argued that training large language models like ChatGPT requires copyrighted content.
“Since copyright covers nearly all forms of human expression—blog posts, photos, forum discussions, code snippets, and government records—it’s impossible to train today’s top AI models without copyrighted material,” OpenAI explained in its submission, as reported by the Telegraph.
“Restricting training data to public domain books and drawings over a century old might be an interesting test, but it wouldn’t produce AI systems that serve today’s needs,” OpenAI added.
Training AI on copyrighted material has become a major debate in tech as generative AI grows. In December 2023, the New York Times sued OpenAI and Microsoft for “unlawful use” of its content to develop their products. In response, OpenAI called training “fair use” and said: “We support journalism, collaborate with news outlets, and believe The New York Times’ lawsuit lacks merit.”
The New York Times’ lawsuit followed a September 2023 lawsuit by 17 authors, including George R. R. Martin of Game of Thrones, who accused OpenAI of “systematic theft on a massive scale.”
In August 2023, District Judge Beryl Howell upheld a U.S. Copyright Office ruling that AI-generated art cannot be copyrighted. The 2018 decision emphasized that “the connection between the human mind and creative expression” is essential for copyright protection.
Latest Downloads
Downlaod
Top News
