Terence Tao Warns 3000 Mathematicians of Dangerous AI Advances

More than 3,000 mathematicians signed the Leiden Declaration last month after Terence Tao warned that AI is advancing faster than the field admits, a threshold moment for a profession whose core work machines are now approaching.

The declaration follows a July 24 appearance in Philadelphia where Tao, widely regarded as the field’s most prominent living figure, told colleagues they are underestimating the speed of AI’s advance, according to Scientific American. “We are not doing a sustainable thing,” Tao said. “AI is going to become extremely capable, and extremely quickly.”

Also Read: Anthropic Researcher Warns of Dangerous AI Risk, Cites 10% Extinction Odds

Separately, Charles Hoskinson, founder of Cardano, said large language models have surpassed his own expectations for mathematical reasoning, while also flagging intellectual property risks tied to how these models are trained and deployed, according to BigGo.

A Field Reassessing Its Own Value

Tao’s warning lands as AI models increasingly handle proof verification and pattern discovery, tasks long considered the exclusive terrain of trained mathematicians. The concern voiced at the Philadelphia gathering was not that AI would replace mathematicians outright, but that the field risks losing its bearings if it treats rapid capability gains as a distant hypothetical rather than a near-term reality.

IP Risk Enters The Crypto Conversation

Hoskinson’s comments add a second dimension to the debate. His IP concerns point to unresolved questions about who owns mathematical insights generated by AI systems trained on public and proprietary research, a question with direct relevance to Cardano’s own use of formal verification methods in blockchain development.

Where The Leiden Declaration Goes From Here

Nature has published guidance urging mathematicians to act swiftly to mitigate AI’s impact on the field, signaling that the Leiden Declaration is an opening move rather than a settled framework.

Read Next: Anthropic CEO Warning of Dangerous AI Agent Swarms in Six Months