Claude just solved a 26 year old cryptography mystery in under 30 minutes. And the answer could change how we think about AI in security research.
Anthropic says Claude traced the origins of the supposedly random seeds used for elliptic curves such as P‑256 to simple timestamps from 1997 and 1998.
NIST published the curves in 2000, but the origins of those parameters had remained unexplained for decades.
That uncertainty helped fuel speculation about whether the NSA could have hidden a backdoor in the curves.
Claude’s discovery offers a much simpler explanation. The $36,864 bounty is set to be donated to the Internet Security Research Group.
But there’s an important distinction here. This does not mean Claude broke P‑256 or cracked modern encryption. Anthropic says the finding makes deliberate backdoor theories much less plausible.
Personally, I think the bigger story is the research capability itself.
AI isn't just generating code or answering technical questions. It can help investigate decades old problems by finding patterns, testing hypotheses and connecting evidence that researchers may have struggled to piece together.
That capability can help defenders audit the systems we already trust. It could also help attackers search for weaknesses faster.
The real question is how many long standing security mysteries AI will help solve next.
Because the same technology that helps us understand our defenses could also help someone find a way around them. 👀
$BTC #BTC Price Analysis# $ETH #Altcoin Season#
Anthropic says Claude traced the origins of the supposedly random seeds used for elliptic curves such as P‑256 to simple timestamps from 1997 and 1998.
NIST published the curves in 2000, but the origins of those parameters had remained unexplained for decades.
That uncertainty helped fuel speculation about whether the NSA could have hidden a backdoor in the curves.
Claude’s discovery offers a much simpler explanation. The $36,864 bounty is set to be donated to the Internet Security Research Group.
But there’s an important distinction here. This does not mean Claude broke P‑256 or cracked modern encryption. Anthropic says the finding makes deliberate backdoor theories much less plausible.
Personally, I think the bigger story is the research capability itself.
AI isn't just generating code or answering technical questions. It can help investigate decades old problems by finding patterns, testing hypotheses and connecting evidence that researchers may have struggled to piece together.
That capability can help defenders audit the systems we already trust. It could also help attackers search for weaknesses faster.
The real question is how many long standing security mysteries AI will help solve next.
Because the same technology that helps us understand our defenses could also help someone find a way around them. 👀
$BTC #BTC Price Analysis# $ETH #Altcoin Season#