AI and the Future of Cybersecurity: Why Openness Matters

人工智能与网络安全的未来：为何开放至关重要

Following the announcement of Mythos and Project Glasswing, institutions throughout the world are grappling with the potential dawn of a new era of cybersecurity. In this post, we break down the current situation, discuss the role of openness, and situate the future of cybersecurity within the larger AI ecosystem. 随着 Mythos 和 Project Glasswing 的发布，全球各地的机构都在努力应对网络安全新时代可能到来的挑战。在本文中，我们将剖析当前局势，探讨开放的作用，并将网络安全的未来置于更广阔的人工智能生态系统中进行考量。

What is Mythos?

什么是 Mythos？

Mythos is a “frontier AI model”, a large language model (LLM) that can be used to process software code (among many other things). This follows a general trend in LLM development, where LLM performance on code-related tasks has recently skyrocketed. What’s particularly significant about Mythos is the system it’s embedded within: It’s the system, not the model alone, that has enabled Mythos to rapidly find and patch software vulnerabilities. Understanding this distinction is key to understanding the current landscape of AI cybersecurity. Mythos 是一个“前沿人工智能模型”，即一种可以用于处理软件代码（以及其他许多任务）的大型语言模型（LLM）。这遵循了 LLM 发展的总体趋势，即近期 LLM 在代码相关任务上的表现呈指数级增长。Mythos 的特别之处在于它所嵌入的系统：正是系统而非模型本身，使得 Mythos 能够快速发现并修补软件漏洞。理解这一区别是理解当前人工智能网络安全格局的关键。

What Mythos demonstrates is that the following system recipe is powerful: substantial compute power; models trained on troves of software-relevant data; scaffolding built to handle software vulnerability probing and patching; speed (enabled by compute power and the capital behind it); and some degree of system autonomy. Together, these ingredients can uncover software vulnerabilities, find exploits, and build patches. It’s in this recipe — not in any one model — that both the benefits and the risks come in. Mythos 展示了以下系统配方具有强大的威力：强大的计算能力；在海量软件相关数据上训练的模型；用于处理软件漏洞探测和修补的架构；速度（由计算能力及其背后的资本支持）；以及一定程度的系统自主性。这些要素结合在一起，可以发现软件漏洞、寻找攻击路径并构建补丁。正是这种配方——而非任何单一模型——带来了机遇与风险。

How Openness Can Be a Structural Advantage

开放如何成为一种结构性优势

As autonomous systems that identify software vulnerabilities proliferate (and they will), open code and tooling can help level the playing field. Software security has become a speed race across four stages: detection, verification, coordination, and patch propagation. Open ecosystems distribute these across a community, where more closed-source projects centralize knowledge and action across all four stages inside a single vendor, representing a single point of failure where only one organization can see and fix the code. 随着能够识别软件漏洞的自主系统不断普及（这是必然趋势），开放的代码和工具将有助于创造公平的竞争环境。软件安全已演变为一场跨越四个阶段的速度竞赛：检测、验证、协调和补丁分发。开放生态系统将这些任务分配给整个社区，而封闭源代码项目则将所有四个阶段的知识和行动集中在单一供应商内部，这构成了一个单点故障，即只有一家机构能够查看并修复代码。

The distributed nature of open development is robust to such constraints, and can be especially powerful in communities with dedicated security professionals, like the Linux kernel security team, the Open Source Security Foundation, and the team at Hugging Face working on model and supply-chain security. 开放式开发的分布式特性对上述限制具有很强的鲁棒性，在拥有专业安全人员的社区中尤为强大，例如 Linux 内核安全团队、开源安全基金会（Open Source Security Foundation）以及 Hugging Face 致力于模型和供应链安全的团队。

Building Defenses with Open Tools and Semi-Autonomous Agents

利用开源工具和半自主智能体构建防御

Cybersecurity defense is where open source and AI agents can play a key role together. Based on the System Card, it appears that Mythos is capable of operating with close to full autonomy, something we’ve advised against due to the potential loss of control. AI agents that are instead semi-autonomous, where the types of actions they can take are prespecified and certain steps require human approval, hit a sweet spot of benefit and risk. 网络安全防御正是开源和人工智能智能体可以共同发挥关键作用的领域。根据系统卡（System Card）显示，Mythos 似乎具备近乎完全自主的操作能力，但由于存在失控风险，我们并不建议这样做。相比之下，半自主的人工智能智能体——即其可采取的行动类型是预先指定的，且某些步骤需要人工批准——在收益和风险之间找到了一个最佳平衡点。

In semi-autonomous systems, people remain in control, and the AI agent is responsible for specific subtasks. This is possible to do with open code that organizations can run privately within their own institutions, specifying allowable tools, skills, and system access privileges. With this setup, AI agents can be deployed defensively, finding vulnerabilities and assisting with patching under an organization’s own controls. 在半自主系统中，人类保持控制权，而人工智能智能体负责特定的子任务。通过开源代码，机构可以在其内部私有运行，并指定允许使用的工具、技能和系统访问权限，从而实现这一点。通过这种设置，人工智能智能体可以被部署用于防御，在机构自身的控制下发现漏洞并协助修补。