AI Solves 25-Year-Old Internet Security Mystery

For 25 years, a crucial piece of internet security—called ECDSA—has had a nagging question: where did its random numbers come from? These numbers, called 'seeds,' are used to create the locks that protect your online banking, shopping, and private messages. The NSA provided the seeds in 2000 without explanation, sparking fears they might contain a secret backdoor for spying.

Now, an AI model named Claude, made by Anthropic, has cracked the case. In under 30 minutes, Claude discovered that the seeds weren't from a complex hash or a secret formula. Instead, they were generated by a common random-number generator from the 1990s, seeded with a simple Unix timestamp—the number of seconds since January 1, 1970. Essentially, the seeds were just based on the time they were created.

This is good news. It means the seeds were generated in a straightforward, reproducible way, making it very unlikely they were deliberately weakened. The discovery increases trust in ECDSA, which secures trillions of dollars in transactions daily. The AI's method was clever: it recognized patterns in the seeds that matched the output of an old C library's random function.

So, what does this mean for you? Nothing changes—you don't need to update anything. But you can rest a little easier knowing that one of the pillars of internet security is probably free of government backdoors. And it's a reminder that AI can solve puzzles that stumped humans for decades, which could lead to more secure technology in the future.