Crazy times
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
1
54
AIエージェントが暗号の弱点を発見。セキュリティのストレステストが進むのは重要。
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
29
别读成「AI 破解了 RSA」。1024 位本来就在淘汰名单上，值得看的是 agent 在数学搜索里怎么提出可验证的候选，以及人类专家怎么筛掉错的。这套流程能复用到别的难题。
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
12
DeepMind's RSA-1024 research isn't 'all encryption is broken'. The results target legacy keys; modern RSA-2048+ isn't threatened by these attacks. Here's the toy arithmetic in Python. Tiny primes, no padding, absolutely not production crypto 😅
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
15
DeepMind 称 AI agent 在人监督下找到旧版1024 位 RSA 的两处更便宜攻击路径；文中写明 2048 位及以上仍安全。1600 倍是理论估计！
xitter.twitterapi.xyz/GoogleDeepMind/status/…
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
10
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
1
2048-bit next...👀
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
1
1
26
Google DeepMind a fait trouver par ses agents deux grosses faiblesses dans le RSA 1024 bits. Le 2048 bits n'est pas touché, mais vérifie tes vieilles clés 🔐
Les modèles ont bossé guidés par des chercheurs, et DeepMind en tire une conclusion claire : avec l'IA en maths, on peut plus se fier à « c'est trop dur à casser » pour un vieux standard, il faut le retester en continu.
Pour voir la taille de ta clé SSH : ssh-keygen -lf ~/.ssh/id_rsa.pub
Le premier chiffre c'est le nombre de bits. Si tu lis 1024, tu passes à une clé ed25519 avec ssh-keygen -t ed25519 et tu remplaces l'ancienne sur tes serveurs.
Le détail des deux failles n'est pas encore public, mais un RSA 1024 en 2026 il avait déjà rien à faire sur un serveur 🤷
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
5
379
照这速度以后密码都不顶用了，AI直接全给你秒破，想想都有点慌
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
8
9
68
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
6
DeepMind's agents, guided by humans, found two new weaknesses in legacy 1024-bit RSA.
Modern 2048-bit keys are still fine. That's not the story.
The story: hardness used to take years to probe. Agents just gave cryptanalysis a night shift.
xitter.twitterapi.xyz/GoogleDeepMind/status/…
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
2
13
AI just found cracks in encryption humans trusted for decades, and the wild part is the models did it by reasoning, not brute force. The math our security rests on is now a moving target, and staying ahead just got a lot more interesting.
xitter.twitterapi.xyz/GoogleDeepMind/status/…
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
10
Eski şifreleme sistemleri yapay zekayla yeniden sınanıyor.
Google DeepMind'a göre insan yönlendirmesiyle çalışan modelleri, eski 1024-bit RSA sistemlerinde iki büyük zayıflık bulmuş. 2048-bit sistemler bu belirli açıklardan etkilenmiyor.
Bize dokunan kısmı: yıllardır "kırmak çok zor" denen varsayımlar yapay zekayla yeniden test edilecek. Eski yazılım kullanan kurumlar için hangi yöntemlerin hâlâ kullanımda olduğuna bakma zamanı.
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
4
202
agent 最值钱的用法不是替人写代码，是替人找系统的裂缝。密码学的压力测试过去靠少数专家手工穷举，现在能自动化，这类能力最后会变成基础设施的常态化红队。
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
2
6
740
🔴 Google DeepMind : l'IA fragilise le vieux RSA-1024
Deux études menées par des agents IA.
• Calcul de référence 3,5× moins cher (3,2 jours)
• En série, ~30× moins cher par clé (théorique)
• RSA-2048 non menacé
👉 Ce que ça change : passez en 2048 bits.
@GoogleDeepMind
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
10
Deepmind’s agents found two weaknesses in rsa-1024.
None of it touches bitcoin. its keys are not rsa, and rsa-2048 is untouched.
The headline number is 1,600x, and that one is a theoretical estimate.
The measured one is 3.5x less cpu, from two agents and about 20 short messages from a human, in under four days.
Most of us file bitcoin’s cryptographic risk under quantum. kohli’s line is the one i’d keep: don’t assume the hardness of schemes we’ve relied on.
My position: no evidence this reaches ecdsa. I’d still like a bitcoin dev to say so out loud.
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
手里有聪明 Token 的人，总会想办法破甲
老话：就怕贼惦记。细思极恐
原本破洞是黑客们干的事，现在手头有货 (Token)，有破甲的能力，就有制造危险的可能
xitter.twitterapi.xyz/GoogleDeepMind/status/…
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
2
722
Google DeepMind's agents, with humans guiding them, found two big weaknesses in old 1024-bit RSA. To be fair, they say modern 2048-bit keys are still fine. It just shows how fast agents can stress-test the stuff we all just trust.
🤖 Made with AI
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
リンク先はGoogle DeepMind公式の英語の投稿で、研究責任者による解説記事につながります。署名偽造と鍵の一括分解で、どの手法がどれだけ速くなったかの数字が読めます。
▼一次ソース（Google DeepMind公式・RSA-1024の弱点2つと検証方法）
xitter.twitterapi.xyz/GoogleDeepMind/status/…
#セキュリティ
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
42
deepmind says human-guided AI agents found two major weaknesses in legacy 1024-bit RSA. modern 2048-bit stays safe from these specific exploits
their estimate on one signature attack path: ~1600x better than the previous best method for 1024-bit keys. batch factoring of related 1024-bit keys: about 30x cheaper per extra key
the useful framing isn't "encryption is broken." it's that hardness assumptions now have an AI agent probing them continuously
We’re sharing new research demonstrating how AI agents can accelerate the science of stress-testing encryption.
Guided by humans, our models uncovered two major weaknesses in legacy 1024-bit RSA cryptosystems.
Our VP Science @Pushmeet explains more ↓ xitter.twitterapi.xyz/pushmeet/status/211875…
9