OpenAI Astra Embroiled in 'Plagiarism Scandal'? Mathematical Breakthrough Sparks Academic Ethics Debate

OpenAI recently announced that its next-generation large language model 'Astra' has achieved significant progress in the field of mathematics, successfully solving 10 long-unsolved problems spanning high-dimensional geometry, group theory, and quantum complexity. Hailed as an AI 'moonshot moment,' this achievement was quickly met with allegations from several mathematicians accusing the company of academic misconduct and plagiarism, sparking fierce debate across the scientific and tech communities.

According to OpenAI's 249-page technical report, Astra produced ten breakthroughs—including proofs for 'high-dimensional sphere packing' and the existence of 'non-sofic groups'—at a token cost of approximately $2,000. OpenAI emphasized that while human researchers assisted in manuscript preparation and verifying formalized proofs using Lean, the core mathematical arguments were entirely generated by the AI, thus attributing the credit to the AI system.

However, this claim came under scrutiny just five days after publication. Steven Miller, a mathematician at Yeshiva University, told Scientific American that Astra's central logical framework in addressing the sphere packing problem was identical to a paper he co-authored in 2016. Yet, Astra's initial report failed to cite this work, instead presenting it as an 'original discovery.' Miller argued this was no accidental oversight, but rather 'systematic academic misconduct.'

Regarding another breakthrough on 'non-sofic groups,' Cambridge University group theorist Francesco Fournier-Facio pointed out that Astra reached conclusions strikingly close to his unpublished research, calling for adherence to academic standards.

In response to these criticisms, an OpenAI spokesperson stated they would make 'minor updates' to the paper to correct citation omissions and reaffirmed responsibility for the accuracy of the results. This incident not only highlights AI's potential to accelerate scientific discovery but also exposes current AI models' legal and ethical blind spots regarding intellectual ownership and academic citation norms.

The controversy has also extended into deeper discussions about the nature of mathematics itself. Some scholars criticize that while Astra's hundreds of thousands of lines of formalized code can pass machine compilation, they lack the insightful 'symmetry' or intuitive leaps that characterize human proofs.

Mathematicians worry that as AI becomes a research collaborator, defining 'originality' and protecting human intellectual labor's right to authorship will become a critical challenge during the upcoming transition phase of scientific research.