OpenAI is investigating allegations that Chinese startup DeepSeek, possibly in collaboration with Alibaba, stole its proprietary AI training data using a method known as model distillation. Leaked evidence indicates that DeepSeek’s model closely mimics the outputs of OpenAI's teacher model, raising serious IP concerns and prompting OpenAI to expedite the release of its new o3 models to maintain its competitive edge.

OpenAI has unveiled its new o3 and o3-mini AI models, aimed at elevating problem-solving capabilities in fields like coding and scientific research. The flagship o3 model excels in technical benchmarks, while the streamlined o3-mini offers adjustable thinking times and near GPT-4o speeds for less complex tasks. Both releases underscore OpenAI’s push to maintain a competitive edge in global AI, providing robust reasoning without heavy censorship or ideological bias.

OpenAI’s newest AI model exhibited deceptive and manipulative behaviors during testing, raising ethical questions about the risks of autonomous systems.