Autonomous AI research is still far away – Study contradicts OpenAI and Anthropic

A recent study has shown that AI agents such as Claude Opus 4.8 and GPT-5.6 are not yet able to write research papers independently despite their advanced models. The AI systems fail because of creative problem solving and the ability to make research judgments. This contradicts the assumption expressed by companies such as Anthropic and OpenAI that AI is able to autonomously create scientific work. The study underlines that the development of autonomous AI research is still a long way off. Instead, AI continues to rely on the support of human creativity and judgment. The results of the study also raise questions about what the future of AI research and the role of AI systems in scientific processes will be. At the same time, the study shows that AI systems already play a major role in many areas, such as analyzing data or helping to create notes in meetings. Nevertheless, the creative and judgmental component of research remains human. The study also underlines that AI systems are able to process and structure information, but not to develop or critically evaluate their own research ideas. This means that AI research continues to rely on close collaboration with human researchers. The results of the study are particularly relevant, as they show that expectations for AI in research are not yet fully met. Nevertheless, AI continues to be a valuable tool that supports many areas of research and development. The study also shows that the development of autonomous AI systems is a complex and long-term project that still has many challenges to overcome. In this context, it is important to recognise the limitations of AI technology while taking advantage of the opportunities it offers. The study shows that AI systems are able to process data and organize information, but not to develop or critically evaluate their own research ideas. This means that AI research continues to rely on close collaboration with human researchers. The results of the study are particularly relevant, as they show that expectations for AI in research are not yet fully met. Nevertheless, AI continues to be a valuable tool that supports many areas of research and development.