OpenAI Astra 与 Claude Opus 5 破解两则长期未解的 Enigma 密文

计算机先驱艾伦·图灵：https://en.wikipedia.org/wiki/Alan_Turing 最广为人知的是他同名的实验，用以测试人工智能和人类智能是否可以区分。然而，他一生中面对的更重要的考验是破译纳粹德国在二战期间使用的恩尼格码。

现在，一对密码分析师表示，他们已经使用 OpenAI 和 Anthropic 构建的大型语言模型（LLM）破解了两条长期未解的恩尼格码机器加密消息：https://en.wikipedia.org/wiki/Enigma_machine。

尽管图灵和他的密码分析团队制造了一台早期计算机，称为 Bombe：https://en.wikipedia.org/wiki/Bombe，使英国在战争期间能够翻译恩尼格码消息，但仍有少数档案消息未被破译——通常由于原编码者的误抄或错误。

开发者卡特·莱芬（Carter Leffen）只是简单地告诉 OpenAI 最新模型 Astra 在恩尼格码消息数据库中搜索一条未被破译的消息并进行解码。该模型正是完成了这一任务：https://x.com/carterleffen/status/2100405732935868840，过程中它进行了自己的档案研究，寻找上下文线索，构建了恩尼格码模拟器，并最终恢复了自 2005 年以来一直让研究人员困惑的明文消息：https://en.wikipedia.org/wiki/Plaintext。他甚至使用 Astra 构建了一个互动网站：https://mvueh-enigma-solved.carterl.chatgpt.site/，解释整个问题。

弗罗德·韦里鲁德（Frode Weirerud）是一位热衷密码学的退休电气工程师，他维护的网站 Crypto Cellar：https://cryptocellar.org/ 包含各种资源和记录，以及一条消息数据库。上周，韦里鲁德验证了：https://www.cryptocellar.org/bgac/the-mvueh-break.html 莱芬的解决方案，他表示对此感到“敬畏”。

值得注意的是，鉴于 AI 代理在回答面前问题时会付出的努力，Astra 模型的日志包括讨论位于“私人收藏”中的档案消息，这些消息并未由韦里鲁德托管。他仍不确定模型是否访问了这些消息，但推测它们可能由其他研究人员在某处网上分享，或者模型能够访问德国政府的公共档案。

“GPT–6 Astra表现得像非常专业的密码分析师和档案研究员，”他写道。“它在两天内完成的成就，人类研究者可能需要数周甚至数月。就我个人而言，我花了数周时间研究GPT–6 Astra所指的联邦档案馆文件。”

9月21日，另一位密码分析师杰克·威利斯联系韦雷鲁德，称他使用了Anthropic的Claude Opus 5模型破解了另一条未破解的消息：https：//cryptocellar.org/bgac/the-fmngi-break.html。威利斯为克劳德提供了更多指导，最终克劳德利用某名官员姓名的已知签名破译该消息。

Weirerud指出，目前仅剩七条未破解的恩尼格码消息，以及一条明文已知但代码未破的消息。

通过我们文章中的链接购买，我们可能会获得少量佣金：https：//techcrunch.com/techcrunch-affiliate-monetization-standards/。这不影响我们的编辑独立性。

你的下一个重要联系是在Disrupt。与10,000+位创始人、风险投资人、运营商和科技领袖建立联系。探索明日的突破，聆听塑造当今科技的动态，并于9月25日太平洋时间晚上11：59前节省最多200美元。

Meta 人工智能代理 Muse 的所有新功能：https://techcrunch.com/2026/09/23/everything-new-coming-to-metas-ai-agent-muse/ Kirsten Korosec：https://techcrunch.com/author/kirsten-korosec/ Lucas Ropek：https://techcrunch.com/author/lucas-ropek/

Meta 为其 Muse AI 代理制作了一个类似电子宠物的可穿戴设备：https://techcrunch.com/2026/09/23/meta-made-a-tamagotchi-like-wearable-for-its-muse-ai-agent/ Lucas Ropek：https://techcrunch.com/author/lucas-ropek/

Anthropic 表示其生物实验室已经发现了重大成果：https://techcrunch.com/2026/09/23/anthropic-says-its-biology-lab-has-already-found-something-big/ Julie Bort：https://techcrunch.com/author/julie-bort/

PitPro 的首台换轮胎机器人在加拿大上线：https://techcrunch.com/2026/09/23/pitpros-first-tire-changing-robot-goes-live-in-canada/ Sean O'Kane：https://techcrunch.com/author/sean-okane/

Anthropic 发布 Opus 5.5，价格更低，性能达到 Fable 级别：https://techcrunch.com/2026/09/22/anthropic-releases-opus-5-5-with-lower-prices-and-fable-level-performance/ Russell Brandom：https://techcrunch.com/author/russell-brandom/

OpenAI 成立数学顾问小组，其 AI 已解决 100 多个未解问题：https://techcrunch.com/2026/09/21/openai-forms-math-advisory-group-as-its-ai-resolves-more-than-100-open-problems/ Aditya Mehta：https://techcrunch.com/author/aditya-mehta/

Meta 的 Muse 超越了 ChatGPT 的早期移动版本表现：https://techcrunch.com/2026/09/21/metas-muse-is-outpacing-chatgpts-early-mobile-launch/ Sarah Perez：https://techcrunch.com/author/sarah-perez/

Computer pioneer Alan Turing：https://en.wikipedia.org/wiki/Alan_Turing is best known for his eponymous experiment to test whether artificial and human intelligence can be distinguished. However, the more important test he faced during his lifetime was cracking the Enigma code used by Nazi Germany during World War II.

Now, a pair of cryptanalysts say they’ve cracked two long-unsolved messages encoded by Enigma machines：https://en.wikipedia.org/wiki/Enigma_machine using LLMs built by OpenAI and Anthropic.

While Turing and his team of cryptanalysts built an early computer, dubbed the Bombe：https://en.wikipedia.org/wiki/Bombe, that allowed the United Kingdom to translate Enigma messages during the war, a handful of archival messages remain unbroken – usually due to mistranscription or errors by the original encoders.

Carter Leffen, a developer, simply told OpenAI’s newest model, Astra, to search a database of Enigma messages for an unbroken message and decode it. The model was able to do just that：https://x.com/carterleffen/status/2100405732935868840, after doing its own archival research, finding context clues, building a simulator of the Enigma machine, and ultimately recovering the plaintext：https://en.wikipedia.org/wiki/Plaintext of a message that had baffled researchers since 2005. He even used Astra to build an interactive website：https://mvueh-enigma-solved.carterl.chatgpt.site/ explaining the whole problem.

Frode Weirerud, a retired electrical engineer with a lifelong interest in cryptology, maintains the website Crypto Cellar,：https://cryptocellar.org/ which includes a variety of resources and records, and a database of messages. Last week, Weirerud validated：https://www.cryptocellar.org/bgac/the-mvueh-break.html Leffen’s solution, which he said left him in “awe.”

Notably, given the lengths that AI agents will go to answer the questions in front of them, the Astra model’s logs include discussion of archived messages in a “private collection” that aren’t hosted by Weirerud. He still isn’t sure if the model accessed them or not, but speculates they may have been shared by a different researcher somewhere online, or that the model was able to access the German government’s public archives.

“GPT–6 Astra is behaving like a very professional cryptanalyst and archive researcher,” he wrote. “What it has achieved in two days would take a human researcher weeks or even months. Personally, I spent several weeks researching the Bundesarchiv files GPT–6 Astra refers to.”

On September 21, another cryptanalyst, Jack Willis, reached out to Weirerud, saying he had used Anthropic’s Claude Opus 5 model to break：https://cryptocellar.org/bgac/the-fmngi-break.html a different unsolved message. Willis provided significantly more guidance to Claude, which was ultimately able to use the known signature of a particular officer’s name to break the message.

Weirerud notes that there are just seven unbroken Enigma messages remaining, along with one message where the plaintext is known but the code is still unbroken.

When you purchase through links in our articles, we may earn a small commission：https://techcrunch.com/techcrunch-affiliate-monetization-standards/. This doesn’t affect our editorial independence.

Your next big connection is at Disrupt. Connect with 10,000+ founders, VCs, operators, and tech leaders. Explore tomorrow’s breakthroughs, hear what’s shaping tech today, and save up to $200 by Sept. 25 at 11:59 p.m. PT.

Everything new coming to Meta’s AI agent Muse：https://techcrunch.com/2026/09/23/everything-new-coming-to-metas-ai-agent-muse/ Kirsten Korosec：https://techcrunch.com/author/kirsten-korosec/ Lucas Ropek：https://techcrunch.com/author/lucas-ropek/

Meta made a Tamagotchi-like wearable for its Muse AI agent：https://techcrunch.com/2026/09/23/meta-made-a-tamagotchi-like-wearable-for-its-muse-ai-agent/ Lucas Ropek：https://techcrunch.com/author/lucas-ropek/

Anthropic says its biology lab has already found something big：https://techcrunch.com/2026/09/23/anthropic-says-its-biology-lab-has-already-found-something-big/ Julie Bort：https://techcrunch.com/author/julie-bort/

PitPro’s first tire-changing robot goes live in Canada：https://techcrunch.com/2026/09/23/pitpros-first-tire-changing-robot-goes-live-in-canada/ Sean O'Kane：https://techcrunch.com/author/sean-okane/

Anthropic releases Opus 5.5 with lower prices and Fable-level performance：https://techcrunch.com/2026/09/22/anthropic-releases-opus-5-5-with-lower-prices-and-fable-level-performance/ Russell Brandom：https://techcrunch.com/author/russell-brandom/

OpenAI forms math advisory group as its AI resolves more than 100 open problems：https://techcrunch.com/2026/09/21/openai-forms-math-advisory-group-as-its-ai-resolves-more-than-100-open-problems/ Aditya Mehta：https://techcrunch.com/author/aditya-mehta/

Meta’s Muse is outpacing ChatGPT’s early mobile launch：https://techcrunch.com/2026/09/21/metas-muse-is-outpacing-chatgpts-early-mobile-launch/ Sarah Perez：https://techcrunch.com/author/sarah-perez/