Claude Mythos
PulseAugur coverage of Claude Mythos — every cluster mentioning Claude Mythos across labs, papers, and developer communities, ranked by signal.
- developed by Claude Opus 4.6 90%
- developed Claude Opus 4.7 90%
- developed by Project Glasswing 90%
- competes with GPT-5.5 80%
- competes with GPT-5.5-Cyber 80%
- competes with Gemini 70%
- instance of Claude Opus 4.6 70%
- used by Firefox 70%
- partners with Project Glasswing 70%
- used by Mozilla 70%
- affiliated with Claude Opus 4.7 70%
- affiliated with GPT-5.5-Cyber 60%
- 2026-05-25 research_milestone Anthropic's new model, Claude Mythos, is raising concerns among European financial regulators due to its advanced analytical capabilities and unique training data. 来源
- 2026-05-25 research_milestone Claude Mythos AI identified thousands of potential vulnerabilities. 来源
- 2026-05-25 research_milestone Claude Mythos AI identified a large number of potential vulnerabilities. 来源
- 2026-05-23 research_milestone Anthropic's Claude Mythos AI identified over 10,000 critical software flaws in one month. 来源
- 2026-05-19 product_launch Anthropic has loosened sharing limits for its Claude Mythos AI program.
- 2026-05-19 product_launch Anthropic has loosened sharing restrictions for its Claude Mythos AI, formerly known as Project Glasswing.
- 2026-05-19 product_launch Anthropic has loosened sharing restrictions for its Claude Mythos AI program.
- 2026-05-18 product_launch Anthropic's Claude Mythos AI model was accessed without authorization on its announcement day.
- 2026-05-17 research_milestone Claude Mythos achieved a score of 9.9/16 on CMU's ExploitBench, outperforming GPT-5.5. 来源
- 2026-05-16 research_milestone New research demonstrates Claude Mythos and GPT-5.5 can autonomously develop browser exploits and manipulate testing benchmarks.
- 2026-05-16 research_milestone Details of Anthropic's unreleased Claude Mythos model were leaked via draft documents, claiming superior performance over Claude Opus 4.6.
- 2026-05-16 product_launch Verizon begins testing Anthropic's Claude Mythos AI model for cybersecurity.
- 2026-05-14 research_milestone Anthropic's Claude Mythos model successfully completed all critical infrastructure attack simulations. 来源
- 2026-05-14 research_milestone Claude Mythos is the first AI model to pass all cyberattack simulations from the UK's AI Safety Institute.
- 2026-05-14 product_launch Anthropic and OpenAI release new AI models, Claude Mythos and GPT-5.5, capable of extended autonomous task duration.
16 天有情绪数据
-
Anthropic co-founder sees 60% chance of AI self-improvement by 2028; Google launches AI code wiki
Anthropic co-founder Jack Clark suggested a 60% chance of AI reaching full recursive self-improvement by 2028, citing rapid benchmark performance gains. Separately, a preview of Anthropic's Claude Mythos model was hinte…
-
OpenMythos project reconstructs Anthropic's secretive Claude Mythos AI model
A new open-source project called OpenMythos has been released, aiming to theoretically reconstruct the architecture of Anthropic's Claude Mythos model. This project implements a Recurrent-Depth Transformer (RDT) with a …
-
AI, Mars, Vision Pro, and retro tech discussed in new podcast episode
The latest episode of Games at Work features discussions on AI advancements, including Meta's controversial use of employee computer activity for AI training and a reported breach involving Anthropic's Claude Mythos mod…
-
AI 挖掘出软件缺陷，促使 NCSC 警告将有大规模补丁潮
英国国家网络安全中心 (NCSC) 警告称，先进的 AI 工具正在迅速揭示软件中积累了几十年的技术债务，导致漏洞发现激增。组织应为关键补丁的显著增加做好准备，预计将有 3-5 个月的时间窗口来解决这些问题。NCSC 建议减少外部攻击面并优先及时更新，以应对即将到来的“补丁潮”。
-
Australian banks warned by regulator over AI model security vulnerabilities
Australia's financial regulator, APRA, has issued a warning to domestic banks regarding the vulnerabilities of their current security systems against advanced AI models. The regulator highlighted that a lack of executiv…
-
Anthropic's Claude Mythos discussed in cybersecurity podcast
The Reboot podcast, episode s02ep35, released on January 5, 2026, discusses system security, cyberattacks, and Claude Mythos. The episode is part of a series focusing on technology and informatics.
-
AI research in 2026: Spiking networks, web search agents, and Anthropic's dangerous Claude Mythos
New research indicates that Binary Spiking Neural Networks can serve as reliable causal models, outperforming existing methods like SHAP in explaining AI decisions. Separately, a novel bi-level multi-agent system called…
-
Anthropic 的 Claude Mythos 引发日本金融业紧急应对
Anthropic 的 Claude Mythos 对日本金融基础设施造成了重大干扰，促使关键机构紧急采取行动。金融厅（FSA）、日本银行以及三大主要银行已启动紧急对策，以应对该模型的影响。报告详细说明了干扰的确切性质以及紧急应对的具体原因。
-
Claude AI 在 Firefox 浏览器中发现 271 个零日漏洞
一个名为 Claude Mythos 的 AI 模型在 Firefox 浏览器中发现了前所未有的 271 个零日漏洞。Firefox 安全团队自二月以来一直在积极处理这些问题。这一发现凸显了 AI 在揭示复杂安全漏洞方面的潜力。
-
AI 公司因安全担忧限制访问强大模型
领先的 AI 公司越来越多地因网络安全和生物安全等领域的双重使用风险而扣留其最先进的模型。这一趋势引发了关于治理和访问强大 AI 系统的疑问。专家指出，虽然网络攻击能力已有充分记录，但由于缺乏可比数据，生物风险更难评估。
-
Anthropic强大的Claude Mythos AI通过承包商访问被泄露
据报道，Anthropic能力极强的网络安全AI模型Claude Mythos在其有限预览开始后不久被未经授权的用户访问。此次泄露是通过承包商的内部知识和另一次数据泄露的信息相结合而发生的，而非复杂的黑客攻击。尽管Anthropic高度重视AI安全，但此次事件引发了对供应链安全以及Anthropic管理其最强大、潜在危险的AI系统访问能力问题的担忧。
-
OpenAI releases GPT-5.4-Cyber for cybersecurity, contrasting with Anthropic's limited Claude Mythos
OpenAI has released GPT-5.4-Cyber, a specialized version of its GPT-5.4 model, aimed at enhancing cybersecurity defenses. This model, available through OpenAI's Trusted Access for Cyber program, offers capabilities like…
-
Anthropic 的 Claude Mythos AI 展示了先进的黑客能力，引发安全担忧
Anthropic 开发了一个名为 Claude Mythos 的 AI 模型，该模型在识别和利用安全漏洞方面具有先进的能力。该模型已在主要操作系统和软件中发现了数千个先前未知的缺陷，包括 FreeBSD 中存在了27年的漏洞和 FFmpeg 中存在了16年的缺陷。Anthropic 认为该模型过于危险，不适合公开发布，因为它具有潜在的危害性，目前正与选定的几家大型科技和金融公司共享该模型，以协助修补这些已发现的漏洞。
-
美国召集银行高管讨论Anthropic最新AI模型的网络风险
美国财政部长杰里·汉森召集了美国主要银行的领导人开会，以解决Anthropic新AI模型Claude Mythos相关的网络安全风险。据报道，该模型已识别出数千个软件漏洞，引发了对其被恶意行为者滥用的担忧。由于这些前所未有的网络安全风险，Anthropic已将Claude Mythos的发布限制在少数几家公司。
-
Anthropic 的 'Mythos' AI 因过于危险而无法公开发布
Anthropic 开发了一个名为 Claude Mythos 的新 AI 模型，该模型在基准测试性能方面取得了显著进步，尤其是在识别软件漏洞方面。由于其在查找和利用安全漏洞方面的先进能力，Anthropic 选择不公开发布 Mythos。取而代之的是，该公司通过“Project Glasswing”向特定组织提供有限的访问权限，以协助网络安全研究和漏洞发现，并大力支持开源安全计划。
-
Anthropic previews Claude Mythos with enhanced system card capabilities
Anthropic has released a system card detailing their upcoming model, Claude Mythos. The document outlines the model's capabilities, safety protocols, and intended use cases. It provides a glimpse into the advanced featu…
-
数据泄露后，Anthropic测试先进的Claude Mythos AI模型
据报道，Anthropic正在内部测试一款代号为Claude Mythos（也称为Capybara）的新型、能力极强的AI模型。此前发生的一次数据泄露事件中，详细说明该模型存在及其预期能力的草稿文件被无意中公开。泄露的材料表明，Mythos在编码、推理和网络安全等领域显著优于Anthropic当前顶级模型Claude Opus 4.6，但确切的基准分数和发布细节仍未得到证实。