文章摘要
多家科技巨头联合发起"玻璃翼计划",利用Anthropic公司开发的Claude Mythos预览版AI模型来检测关键软件漏洞。该AI已发现数千个高危漏洞,包括所有主流操作系统和浏览器的漏洞。鉴于AI能力快速提升,该计划旨在将这种技术优先用于网络安全防御,防止漏洞被恶意利用。
文章总结
项目Glasswing:为AI时代筑牢关键软件安全防线
核心内容概述
Anthropic联合亚马逊云科技、苹果、博通、思科等12家行业巨头启动"Glasswing计划",旨在运用前沿AI模型Claude Mythos Preview的漏洞检测能力,主动防御全球关键软件的安全威胁。该计划将提供1亿美元算力补贴和400万美元开源安全捐赠,通过AI与人类协作重塑网络安全格局。
关键发现
- AI安全能力突破
Claude Mythos Preview作为未发布的尖端模型,展现出超越顶级人类专家的漏洞检测能力:
- 发现OpenBSD存在27年的远程崩溃漏洞(该系统以高安全性著称)
- 识别FFmpeg中16年未被察觉的漏洞(该代码行曾被测试500万次)
- 自主串联Linux内核多个漏洞实现权限提升
- 行业合作机制
- 40余家关键基础设施组织将使用该模型扫描第一方及开源系统
- 微软实测显示:在CTI-REALM安全基准测试中,Mythos相较前代模型有显著提升
- 思科强调:"AI能力已突破临界点,必须采用新方法加固系统"
- 威胁形势演变
- 漏洞利用窗口从数月缩短至分钟级
- 全球网络犯罪年损失估计达5000亿美元
- 国家背景的黑客攻击威胁民生与军事系统
技术指标对比
| 测试项目 | Mythos Preview | 前代模型Opus 4.6 | |-----------------------|---------------|-----------------| | 漏洞复现准确率 | 83.1% | 66.6% | | SWE-bench编程任务 | 77.8% | 53.4% | | 终端操作基准测试 | 93.9% | 80.8% |
实施规划
- 短期:90天内公开漏洞修复成果及安全实践报告
- 中期:制定AI时代安全新标准(含漏洞披露、供应链安全等7大方向)
- 长期:推动建立第三方监管机构协调公私部门合作
战略意义
该计划标志着网络安全进入AI防御新时代,通过将最先进AI能力优先赋予防御方,为民主国家在AI军备竞赛中建立安全护城河。正如摩根大通CISO所言:"当攻击者即将获得AI加持时,我们必须比他们行动得更快。"
(注:原文中大量技术测试细节、合作伙伴具体评价及图片数据因篇幅限制未完整呈现,核心信息已凝练如上)
评论总结
以下是评论内容的总结:
价格优势
- Mythos Preview的价格比GPT 4.5和GPT 5.4 Pro更便宜。
- 引用:
- "Pricing for Mythos Preview is $25/$125, so cheaper than GPT 4.5 ($75/$150) and GPT 5.4 Pro ($30/$180)"
- "So, $100B+ valuation companies get essentially free access to the frontier tools with disabled guardrails"
安全与伦理承诺
- Anthropic强调其模型的安全性和伦理审查,但部分用户质疑其真实性或动机。
- 引用:
- "It's nice to know that they continue to be committed to advertising how safe and ethical they are."
- "oops, our latest unreleased model is so good at hacking, we're afraid of it! literal skynet!"
漏洞检测能力
- Mythos Preview在检测高危漏洞方面表现出色,但部分用户认为这是营销手段。
- 引用:
- "Mythos Preview has already found thousands of high-severity vulnerabilities"
- "Another Anthropic PR release based on Anthropic’s own research, uncorroborated by any outside source"
模型风险与限制发布
- Anthropic因安全风险未公开发布Mythos,但用户怀疑可能是成本或商业策略原因。
- 引用:
- "We were sufficiently concerned about the potential risks...before deploying an early version"
- "Part of me wonders if they're not releasing it for safety reasons, but just because it's too expensive to serve"
开源与访问权争议
- 用户批评Anthropic仅向大公司提供最先进模型,呼吁开源竞争。
- 引用:
- "So they are only giving access to their smartest model to corporations"
- "We better fucking hope open source wins, because we aren't getting access if it doesn't"
长期影响与行业变革
- 部分用户认为AI漏洞检测可能重塑网络安全和军事信号情报,但也有人呼吁立法规范软件开发。
- 引用:
- "It could also totally reshape military sigint in similar ways"
- "We must demand our representatives in government pass laws requiring software be architected...according to a basic set of industry standard best practices"
技术性能亮点
- Mythos在长上下文性能测试中表现优异。
- 引用:
- "based on the system card it seems like they've cracked it: GraphWalks BFS 256K-1M Mythos 80.0%"
- "Mythos Preview most often opens with a statement about its introspective curiosity toward its own experience"
政治与国防关联
- 部分用户注意到Anthropic与美国政府的合作,认为其技术可能用于国防或网络安全对抗。
- 引用:
- "the US and its allies must maintain a decisive lead in AI technology"
- "As Iran engages in a cyber attack campaign today the timing of this release seems poignant"
总结:评论围绕价格、安全性、漏洞检测能力、模型访问权及行业影响展开,既有对技术进步的认可,也有对商业动机和潜在风险的质疑。