Papalocal
Loading…
← All news
Technology

OpenAI's 'Mathpocalypse' rattles crypto and cybersecurity experts — but no encryption is broken yet

OpenAI released solutions to hundreds of unsolved math problems, triggering warnings from Ethereum researchers and cryptographers about AI's long-term threat to the encryption securing Bitcoin, banking, and the web.

How does this story make you feel?

OpenAI released solutions — full or partial — to hundreds of long-unsolved mathematical problems, documented across hundreds of research papers, and the fallout landed far outside academic circles. Within days, a senior Ethereum researcher was urging the blockchain industry to prepare for 'bunker mode,' a co-founder of Ethereum was warning of a threat more immediate than quantum computers, and computer scientist Scott Aaronson had coined a name for the moment: the 'Mathpocalypse.'

Surely one of the biggest days in mathematical history.— Scott Aaronson, computer scientist

Among the claimed advances were progress on some of mathematics' most celebrated unsolved problems — the Riemann hypothesis, the Birch and Swinnerton-Dyer conjecture, and the Hodge conjecture — contained in hundreds of results across 17 mathematical subfields. A single experimental internal model generated each result from one prompt, with roughly three hours of ChatGPT Pro compute per result. That release came on the heels of a separate, widely noted September claim that 10,000 AI agents had cracked the 90-year-old Navier-Stokes equations in fewer than 90 hours.

None of the results explicitly concerns cryptography. That absence, rather than the results themselves, became one of the most discussed aspects of the release.

Cryptography is a subfield that's extremely conspicuous by its absence.— Scott Aaronson, computer scientist

The gap fuelled speculation — unconfirmed — that government censorship may be suppressing cryptographic breakthroughs because of their potential to damage widely used technologies, including the transport layer security (TLS) protocol that powers HTTPS. If such breakthroughs exist and are eventually made public, the concern is that the cryptocurrency industry could be caught off guard and forced into a rushed, disorganised migration of assets.

Why cryptography is at stake: Public-key cryptography — which secures web browsing, messaging, banking, and cryptocurrency transactions — rests on mathematical problems currently considered too difficult for computers to solve efficiently. In cryptocurrency, a private key signs a transaction; a public key, derived from the private key using elliptic-curve mathematics, verifies it. The process is designed to run only one way: reversing it to recover the private key has been historically infeasible. Bitcoin and Ethereum both rely on a specific scheme called ECDSA. The security of ECDSA rests on an unproven assumption that no faster method exists to derive a private key from a public key. AI-driven mathematical progress raises the question of whether that assumption could eventually fail.

The day after the release, Justin Drake — a senior researcher at the Ethereum Foundation whose expertise spans cryptographic protocol and consensus design — took to X to urge the blockchain industry to begin planning for 'bunker mode,' saying the drop demonstrated that mathematical superintelligence is upon us. Drake's worry centers not on quantum computing, a threat most experts place further in the future, but on the possibility of a classical algorithmic breakthrough: a mathematical shortcut capable of running on conventional hardware and unraveling the elliptic-curve mathematics that safeguards Bitcoin and Ethereum wallets.

The blockchain industry to calmly begin planning for 'bunker mode' as the 'OpenAI drop made it clear that mathematical superintelligence is upon us.'— Justin Drake, senior researcher, Ethereum Foundation

Drake pointed not to any cryptography result in the release, but to adjacent breakthroughs — including progress on the n log(n) bound for integer multiplication and the 3SUM conjecture — as evidence that AI is dismantling long-assumed mathematical barriers faster than the public realises. Ethereum co-founder Vitalik Buterin voiced similar alarm, cautioning that swift AI-driven advances in mathematics may threaten cryptocurrencies more immediately than quantum computing, and that even lattice-based post-quantum cryptography could be compromised without substantial increases to key sizes. Buterin further recommended that cryptocurrency holders store funds in unspent wallet addresses, keeping their public keys hidden from any AI-powered cracking attempts.

Cryptographer Matthew Green cautioned that as AI grows more capable in mathematics, it could eventually produce algorithms able to defeat public-key encryption. Scott Aaronson added that, according to unnamed sources, AI companies have now started, gingerly and discreetly, investigating whether their latest internal models can break important cryptographic protocols and primitives.

Not everyone in the field shares the alarm. Yehuda Lindell, Head of Cryptography at Coinbase, pushed back directly on Drake's stance, stating there is no evidence that ECDSA is going to fail. The broader expert consensus, for now, is that no current cryptographic standard has been broken, and that the fear of a near-term AI threat to cryptography is more extrapolation than demonstrated reality.

There is no evidence that ECDSA is going to fail.— Yehuda Lindell, Head of Cryptography, Coinbase

The mathematical community's reaction has been mixed in a different way. Professor Timothy Gowers likened the speed of AI's advance to watching a 'tsunami' approach from the horizon. Fields Medallist Terence Tao circulated a statement from the Association for Human Mathematics urging researchers to end their collaboration with OpenAI, with the group characterizing the large-scale publication of papers as a 'demonstration of power' rather than genuine scholarship. An independent advisory board that OpenAI had consulted went so far as to ask the company to cease running advanced mathematical problems on proprietary internal models.

A central problem complicating all of these debates is verification. As of the time of reporting, none of the math results have been externally verified by human mathematicians — they have been validated only by Lean, a computer program that checks formalized proofs. The Navier-Stokes solution has not been accepted by the Clay Mathematics Institute, which administers the Millennium Prize Problems. Associate professor Valerio Capraro observed that rigorously checking even a single research paper can take months, leaving human experts unable to keep pace as AI systems churn out mathematical claims.

Observers have also noted a possible commercial dimension to the timing and framing of these releases. Mathematics results, unlike breakthroughs in physics or chemistry, require no wet-lab experiments to produce or announce. The field carries a reputation for difficulty and purity, making mathematical progress a potent signal of model capability — one that frontier labs have incentive to advertise. One analysis framed it this way: commercial incentives mean hype and truth can travel together, but their coexistence does not cancel out the truth, and every claim deserves to be judged on its own merits.

For ordinary users, no immediate action is required. The risk, if it materialises, would unfold over years, not days, and would first require independent verification of results that have not yet cleared that bar. The more pressing concern flagged by experts is systemic: if AI-generated mathematical progress outpaces the cryptographic community's ability to detect threats and migrate to new standards, the window to respond safely could narrow faster than anticipated.

Why it matters — If AI-driven mathematical progress outpaces the cryptographic community's ability to detect and respond to new vulnerabilities, the encryption protecting online banking, messaging, and cryptocurrency could face threats before adequate defences are in place — even if no such threat is imminent today.

⚠ Not yet confirmed

  • AI companies have begun quietly investigating whether their latest internal models can break important cryptographic protocols.
  • Government censorship may be suppressing cryptographic breakthroughs in OpenAI's release due to potential damage to widely used technologies.

Sources differ on Whether OpenAI's math results represent a near-term threat to cryptographic security: The drop made clear that mathematical superintelligence is upon us; blockchain industry should plan for bunker mode. (x.com (Justin Drake, Ethereum Foundation)) vs There is no evidence that ECDSA is going to fail. (x.com (Yehuda Lindell, Coinbase))

Sources differ on Whether OpenAI's large-scale math publication represents genuine scholarship: The Association for Human Mathematics characterised it as a 'demonstration of power' rather than genuine scholarship. (tbsnews.net) vs Scott Aaronson called it 'surely one of the biggest days in mathematical history.' (tbsnews.net)

Reported by dnyuz.com, thetimes.com, independent.co.uk, theweek.com, techradar.com, newspaceeconomy.ca, tbsnews.net

People in this story

Community

Be the first to comment.

Share this story
Text Email