What to Know
- An Anthropic AI model traced five unexplained seeds for NIST elliptic curves supplied by the National Security Agency to lrand48, an ordinary random-number generator found in older C software libraries.
- The model reached the result in less than 30 minutes, resolving a cryptographic question that had lingered for 25 years.
- The seeds were tied to timestamps from 1997 and 1998, reducing the likelihood that the NSA deliberately embedded a hidden weakness in the curves.
- Claude reproduced the sequence behind P-256 and found an exact match for its published seed at the 1,842nd candidate.
- Anthropic said users of the NIST curves do not need to take action, although some questions remain about how certain curve elements were selected.
- Bitcoin was not affected because it uses secp256k1, a separate and simpler elliptic curve with little room to hide a backdoor.
- Bitcoin still faces a separate long-term quantum computing concern, because a sufficiently powerful quantum machine could theoretically work backward from a public key to a secret key.
- The related bounty totaled $36,864 and will go to the Internet Security Research Group, the nonprofit behind Let’s Encrypt.
Claude Resolves a Long-Running Cryptography Puzzle
An Anthropic AI model has solved one of modern cryptography’s most persistent mysteries by tracing unexplained numbers supplied by the National Security Agency in the late 1990s to an ordinary random-number program. The numbers, used as seeds for NIST elliptic curves, had long attracted suspicion because the agency provided them without a clear explanation of how they were chosen.
The discovery matters because elliptic-curve cryptography underpins large parts of internet security and is also central to digital-asset custody. The technology allows users to prove they control a secret key without exposing that key, making it essential to secure communications, financial authentication, and cryptocurrency wallets. When the mathematical foundations of that system appear unexplained, cryptographers naturally ask whether the design contains a hidden weakness.
In this case, the Anthropic model traced five NIST curve seeds to lrand48, a random-number generator used in older C software libraries, with timestamps from 1997 and 1998. The model did so in less than 30 minutes, a striking result given that specialists had pursued the question for 25 years. The outcome does not prove that every design choice around the curves was ideal or fully documented, but it substantially weakens the long-running theory that the seeds were chosen after a secret search for exploitable curves.
Why the NSA Seeds Became Controversial
In 2000, the U.S. National Institute of Standards and Technology adopted a set of elliptic curves supplied by the NSA. Some of the starting values behind those curves, known as seeds, were public, but the process used to generate them was not explained. That omission became the basis for a durable concern: if an agency can choose the mathematical starting point, it might be able to test many possibilities and select one with a weakness known only to itself.
The unease deepened after the Snowden disclosures connected the NSA to efforts to weaken cryptographic standards. One especially damaging episode involved the Dual EC random-number generator, which NIST later removed from its recommendations. Against that backdrop, unexplained NSA-linked curve seeds looked less like a documentation gap and more like a possible attack surface.
Cryptographers Neal Koblitz and Alfred Menezes examined the controversy in their 2015 paper, “A Riddle Wrapped in an Enigma.” Menezes had helped write the original standard and recalled receiving the seeds from NSA mathematician Jerry Solinas in 1997. Security researcher Steve Weis also pursued the question for years, including through public-records requests in 2018. NIST said it had no documentation, while the NSA stopped responding.
Another possible explanation suggested the numbers might have come from a humorous phrase Solinas had forgotten, prompting searches through word combinations. That path did not resolve the mystery. In 2023, cryptography engineer Filippo Valsorda and other contributors offered a $12,288 bounty for an answer, rising to more than $36,000 if the winner donated it to charity. Solinas died earlier that year, leaving researchers to rely on memories, mathematical traces, and the public seeds themselves.
How the AI Model Found the Pattern
The breakthrough came when an Anthropic employee asked an internal research model to investigate the seeds without giving it the older theory that they came from a forgotten phrase. Instead of chasing language-based clues, Claude identified a computational origin. The model linked the seeds to lrand48 and reproduced the sequence behind P-256, a widely used NIST curve.
The exact match arrived at the 1,842nd candidate. In the model’s recorded reasoning, Claude described the result as solving the 25-year-old mystery and questioned whether the discovery was plausible in the real world, noting that the bounty had remained unsolved. The model then recognized that the pattern’s statistical features made the explanation credible, calling it a real discovery.
The significance is practical as well as historical. Anyone can now rerun the program and reproduce the numbers the NSA supplied. That transparency changes the risk assessment. If the seeds came from an ordinary random-number generator and timestamps, then the backdoor theory becomes much less compelling. However, Anthropic’s team still cannot explain why the generation process passed over some usable candidates or how other curve details were selected, so the result does not close every possible question.
Why Predictability Is Safe for Public Seeds but Dangerous for Wallets
The finding also illustrates an important distinction in cryptography: public parameters and secret keys face different security requirements. The NSA’s seeds were public from the start, meaning there was no hidden private value for an attacker to steal. Reconstructing the seed generation process helps explain the origin of the public numbers, but it does not directly expose private data from systems using the curves.
By contrast, the same kind of predictability would be catastrophic if it appeared inside a wallet’s secret key generation. A private key produced with a predictable random-number generator could potentially be recreated by anyone who could estimate when the key was made. In cryptocurrency, that would be enough to sign transactions and drain funds. The lesson for crypto users is not that NIST curves suddenly require emergency action, but that high-quality randomness remains essential whenever private keys are created.
Anthropic said people using the NIST curves do not need to change anything in response to the discovery. That position reflects the difference between explaining a public standard’s origin and finding a working exploit. The model appears to have resolved where the seeds came from, not uncovered a new method to break systems relying on those curves.
Bitcoin Was Never on the NIST Curve List
Bitcoin sits outside this particular controversy because it does not use the NIST curve at the center of the debate. Bitcoin wallets rely on secp256k1, a different elliptic curve defined by a simpler equation. Its structure is often viewed favorably by cryptographers because it leaves little room for hidden parameter manipulation.
The difference is straightforward. Bitcoin’s curve is defined by two small, fixed numbers, while P-256 includes a 77-digit number derived from the unexplained NSA seed. With fewer adjustable parts, secp256k1 offers less apparent room for anyone to quietly embed a weakness. That does not make Bitcoin immune to every possible cryptographic risk, but it does mean the long-running NSA seed mystery did not directly involve Bitcoin’s chosen curve.
Satoshi Nakamoto never publicly explained why secp256k1 was chosen, and the curve was rarely used before Bitcoin made it prominent. Ethereum later adopted secp256k1 as well. Bitcoin also retained it when the 2021 Taproot upgrade introduced a newer signature style, reinforcing its role as a core element of the network’s cryptographic identity.
The Quantum Question Has Not Gone Away
While the NSA curve mystery appears much less threatening after Claude’s discovery, Bitcoin still faces a separate and longer-term issue from quantum computing. A sufficiently powerful quantum computer could theoretically work backward from a public key visible on Bitcoin’s blockchain to the secret key behind it. That risk is not about whether secp256k1 contains a hidden backdoor; it arises from the mathematics of elliptic-curve cryptography itself.
The distinction is important for investors, developers, and policy observers. The NSA seed issue concerned whether a human institution had manipulated public curve parameters. The quantum issue concerns whether future hardware could undermine the hardness assumptions that protect elliptic-curve systems. Solving one mystery does not solve the other.
Progress in quantum hardware continues to draw attention. Google said this week that its Willow chip reduced errors about 40-fold in one of the costliest steps such a machine would need. Even so, no one has built a quantum computer anywhere near large or reliable enough to crack a Bitcoin key, and there is no clear timeline for when such a machine might arrive.
What the Discovery Means for Crypto Markets
For the crypto market, the immediate takeaway is narrower than the headlines may suggest. Bitcoin was not secretly implicated in the NSA curve debate, and the discovery does not create a new vulnerability in Bitcoin wallets. Instead, it clarifies an old question about NIST curve seed generation and underscores why transparent cryptographic design matters.
The result may also strengthen confidence in the broader review culture around cryptography. For years, researchers, engineers, and security specialists kept pressure on an unresolved standards question. The answer ultimately came from an AI model, but only after decades of human scrutiny had framed the problem, preserved the evidence, and funded incentives for a solution.
The bounty connected to the search will go to the Internet Security Research Group, the nonprofit behind Let’s Encrypt, which provides websites with free security certificates. That outcome gives the discovery a fitting close: a mystery about the foundations of online trust is ending with support for an organization dedicated to making secure connections widely available.
AI’s Expanding Role in Security Research
Claude’s role in this discovery is likely to intensify debate about how AI models can assist technical research. The model did not merely summarize a known answer; it identified a plausible computational pathway and produced a reproducible explanation. For cryptography, where small patterns can carry major implications, that capability may become increasingly valuable.
At the same time, the case highlights why human verification remains essential. A model’s reasoning must be checked, rerun, and evaluated against known facts. In this instance, the ability for others to reproduce the sequence is what gives the claim weight. Without reproducibility, even an impressive AI-generated hypothesis would remain just a hypothesis.
For FXCOINZ readers, the broader message is that cryptographic confidence depends on open inspection, clear standards, and disciplined risk separation. The NSA curve question is now more explainable. Bitcoin’s curve choice remains outside that controversy. Quantum computing remains a separate issue to watch, not an immediate break in the network’s security.
Frequently Asked Questions (FAQs)
What did Claude discover about the NSA curve seeds?
Claude traced five unexplained NIST elliptic-curve seeds supplied by the NSA to lrand48, an ordinary random-number generator found in older C software libraries, using timestamps from 1997 and 1998.
How long did the AI model take to solve the mystery?
The Anthropic model reached the result in less than 30 minutes, despite the question having remained unresolved among cryptographers for 25 years.
Why were the NSA-supplied seeds controversial?
The seeds were controversial because the NSA supplied the curve designs without explaining how some numbers were chosen, raising concern that the agency may have selected values that created a hidden weakness.
Does this prove there was no NSA backdoor?
The discovery makes a hidden backdoor much less likely because the seeds can now be reproduced through an ordinary program, but it does not explain every design choice involved in the curves.
Do users of NIST curves need to take action?
Anthropic said people using the NIST curves do not need to change anything, because the discovery explains the public seed generation process rather than revealing a new practical exploit.
Was Bitcoin affected by the NSA curve mystery?
No. Bitcoin uses secp256k1, not the NIST curves involved in the mystery, so the unexplained NSA seeds were not part of Bitcoin’s wallet cryptography.
Why is secp256k1 viewed differently from P-256?
Secp256k1 is defined by a simpler structure with two small, fixed numbers, while P-256 includes a 77-digit number derived from an NSA-linked seed, leaving Bitcoin’s curve with less room for hidden parameter selection.
Does Bitcoin still face a quantum computing risk?
Yes. A sufficiently powerful quantum computer could theoretically derive a secret key from a visible public key on Bitcoin’s blockchain, but no computer anywhere near large or reliable enough to do that has been built.
Where is the bounty money going?
The $36,864 bounty will go to the Internet Security Research Group, the nonprofit behind Let’s Encrypt, which provides websites with free security certificates.