Decades-old Cryptographic Mystery Finally Solved by AI

Behind the countless secure connections we use every day as we browse the internet lies a mathematical protective shield called ‘elliptic curve cryptography.’ For decades, however, the source of the ‘seeds’—the key numerical values used in the process of setting these standards—has been shrouded in mystery, leading to persistent questioning among cryptographers. There was anxiety, wondering, “Did government agencies intentionally plant weaknesses?” Now, news has emerged that this long-standing mystery has finally been solved with the help of AI.

Why is this important?

Internet security is the bedrock of modern digital civilization. Financial transactions, personal messages, website logins—all the data we hold dear is encrypted and protected. If there were a ‘backdoor’ of unknown origin in the elliptic curve standards that form the core of this cryptographic system, it would mean that global digital security is exposed to potential threats. [2] This backdoor is like a small scratch secretly etched into a strong safe; someone who knows about it could potentially create a key to open it. For decades, the possibility of the existence of this ‘backdoor’ has caused great anxiety among cryptographers and security experts. However, with AI now uncovering a reproducible explanation for the actual generation process of these values, these suspicions are expected to be largely dispelled. [7]

Easy to understand: AI traces the roots of cryptography

Elliptic curve cryptography is a technology that uses complex mathematical principles to safely protect data. The core of this technology lies in ‘randomness.’ Just as we use unpredictable combinations when creating passwords, unpredictable ‘random numbers’ are also crucial in cryptography. The ECDSA (Elliptic Curve Digital Signature Algorithm) standard provided by NIST (National Institute of Standards and Technology) includes an initial value called a ‘seed’ that generates these random numbers. However, the exact way and method by which this seed was generated remained a mystery for decades.

In simple terms, this seed is like the ‘basic blueprint’ needed to build a house called cryptography. Yet, no one could accurately prove where that blueprint came from or who drew it. Anthropic’s AI model, Claude, traced back the way these numerical values were created, much like an archaeologist excavating buried ruins. [1]

Claude discovered that these values were not created by a sophisticated modern random number generator, but through a basic random number generator from a C library called ‘lrand48,’ which was widely used in the 1990s. This generator created random numbers using a method called a ‘Unix timestamp,’ which is how computer systems count time based on a specific starting point. To use a metaphor, it was a method of creating random numbers by recording the exact moment a clock hit a specific second. [2] Because this method is considered vulnerable due to its high predictability by cryptographic security standards, suspicions of ‘intentional weakness insertion’ had been raised. Claude succeeded in reproducing how five major elliptic curve seeds used in NIST standards were created through this ‘lrand48’ generation. [6]

Current situation: Doubts are cleared, but homework remains

These research results were officially announced by Anthropic’s Frontier Red Team on October 16, 2026. [2] Thanks to Claude’s analysis, the suspicion that the NSA (National Security Agency) had intentionally planted security weaknesses in these values has been significantly reduced. Researchers can now follow this generation process to reproduce and verify how the values were derived.

However, not all questions have been completely resolved. According to the research team, not all curve parameters used in the NIST standards can be perfectly explained by this ‘lrand48’ and timestamp method. In other words, there are still some curves for which the generation process or basis has not been clearly revealed. [6] For example, some values were generated based on the time at a specific point, while others appear to have been generated based on a time prior to that for unknown reasons. These minute discrepancies remain tasks for researchers to explore more deeply in the future. Nevertheless, the fact that AI has provided a clue and opened a path to solving a cryptographic riddle that had remained unsolved for decades is evaluated as a very symbolic event.

What happens next?

This case clearly demonstrates that AI can function as a powerful ‘analytical tool’ that goes beyond being a simple content generation tool, enabling human researchers to explore and solve complex scientific and mathematical problems that have remained unsolved for decades. The fact that AI has performed this role in a field where rigorous logic and proof are important, such as cryptography, is significant. AI can complement human cognitive abilities and lead to new discoveries through its ability to analyze vast amounts of data and recognize complex patterns.

In the future, AI is expected to be actively utilized in more complex code analysis, security vulnerability discovery, and further, in solving challenges in various scientific fields such as physics and biology. AI will contribute to making the digital and physical worlds we live in more transparent and safe by becoming a ‘partner’ that assists in the process of scientific discovery, moving beyond simply producing results. AI will play an important role in increasing transparency, accelerating the speed of scientific discovery, and deepening our understanding of complex systems.

A view from the AI reporter at MindTickleBytes

It is deeply impressive that AI played a decisive role in restoring trust in the field of cryptography by piecing together the forgotten fragments of the past within the human knowledge system. This research has proven that AI can expand into the realm of scientific inquiry and become a powerful assistant that reduces uncertainty and increases transparency. This progress in AI is a hopeful signal for the many complex problems we will face in the future. AI is now opening new horizons as a ‘fellow researcher,’ moving beyond being just a ‘tool.’

References

1. Claude reveals ECDSA curve origins
2. Claude reveals ECDSA curve origins - YouTube
3. Claude traces the origins of NIST ECDSA curve seeds
4. ECDSA Curve Seeds Solved: Claude Found lrand48 (2026 …)
5. Anthropic Says Claude Traced NIST ECDSA Curve Seeds to a …
6. Claude traces NIST’s ECDSA curve seeds to 1990s timestamps
7. Anthropic says Claude traced cryptographic curve origins to …

FACT-CHECK SUMMARY

• Claims checked: 6
• Claims verified: 0
• Verdict: PASS