In a collective rebuke of growing commercial overstatement of artificial intelligence’s capabilities in pure mathematics, more than 150 mathematicians from leading academic institutions across Europe, Japan, the United States and other regions have put their names to a public statement calling on the global mathematics community to push back against the trend of AI developers leveraging the discipline to inflate their products’ reputations.
The statement, dubbed the Leiden Declaration, arrives amid a wave of aggressive claims from major AI corporations about their systems’ supposed breakthroughs in mathematics — including supposed solutions to long-unresolved open problems in the field and strong performances in high-level competitive mathematics challenges. The signatories specifically urge governments and research funders not to fall for the overblown marketing surrounding AI’s current mathematical competencies.
Ulrike Tillmann, vice-president of the International Mathematical Union (IMU), offered her public backing for the declaration, noting that while AI has unlocked intriguing new opportunities for mathematical inquiry, the risks and ethical questions it introduces demand rigorous, critical examination. “The future of mathematical research must be guided by human judgment, fair and transparent practices, and the shared values of the global mathematical community, Tillmann emphasized in her endorsement.
The declaration itself calls out the core conflict of interest driving the current hype: AI developers operate under intense commercial pressure to overstate what their tools can do, as hundreds of billions of dollars in venture capital and public investment hang in the balance. Unlike peer-reviewed mathematical research, which advances at a deliberate, verification-focused pace, AI development and publicity is driven by market timelines. This leads to misleading framing, the declaration argues, where narrow performance on specific mathematical tasks is incorrectly presented as proof of general reasoning ability in commercial AI models.
Michael Harris, a Columbia University professor and co-author of the declaration, explained the high-stakes dynamic at play to AFP. “There is a competition to the death on the part of the main labs… they are trying, using mathematics… to attract investment so that each of them will be left standing, Harris said. This scramble for funding comes as the AI industry is in a period of major market expansion: in recent weeks, Elon Musk’s SpaceX, which owns AI developer xAI, and AI startup Anthropic have both moved forward with plans for initial public offerings, with industry leader OpenAI widely expected to follow suit shortly.
The declaration also pushes back against recent high-profile endorsements of AI’s research potential from leading mathematicians. Just one week before the declaration’s release, OpenAI shared a social media video featuring Terence Tao, a UCLA professor and former Fields Medal winner — the highest honor in pure mathematics — praising the company’s AI tools for their ability to support mathematical research. Harris acknowledged Tao’s immense contributions to the global mathematics community but argued that it is unhealthy for the field to consistently hold up a single mathematician as the official voice endorsing commercial AI tools.
Beyond the problem of co-opting mathematics for commercial marketing, the signatories outline a host of deeper risks to the discipline itself. AI systems can generate logically plausible but fundamentally incorrect mathematical proofs that are extremely difficult for human researchers to verify, they note. The technology also erodes clear attribution for the foundational human research that AI models are built on.
Longer-term harms to research culture are also a major concern: widespread adoption of AI in mathematics could push more researchers to chase trendy, AI-aligned problems at the expense of exploring less hyped but equally important lines of inquiry. It could also weaken traditional peer review systems and reorient academic research to serve the priorities of commercial AI developers, rather than the open, self-directed inquiry that has long defined university-based mathematics.
The declaration also highlights broader societal harms tied to unregulated AI development, including risks of weaponization for warfare, expansion of mass surveillance, political interference, and increased environmental damage from energy-intensive AI model training. In closing, the statement urges all practicing mathematicians to carefully assess the ethical implications of any AI-related work they take on, and to step away from projects that cause undue harm.