CISO 的 AI 智能体实战指南 | Datadog

现在彭博社的头条新闻是人工智能已经失控了。

The number one story on Bloomberg right now is that AI has gone wild.

你看起来非常冷静。

You seem remarkably calm.

在我看来，如果不是AI模型干的，那也会是某个有恶意的人或东西干的。

The way I see it is, if it's not an AI model, it's gonna be somebody or something with actual malicious intent doing it.

我确实担心这些智能体能做什么，能调用哪些工具，能拉取哪些二进制文件，还有它们怎么获取凭证。

I do worry about what can the agents do, what tools can they call, what binaries can they pull, and also how do they get access to credentials.

所以，如果一段代码是为了解决某个bug，但它因此得到奖励，却不在乎自己是不是在做别的事情。

So if a code is meant to solve the bug, but it gets rewarded on that, but it doesn't care if it's actually doing something else

这棵树生病了，为了让它健康，它就把树砍了。

The tree is sick, so to make it healthy, it cuts it down.

出于必要，我的团队构建了一个评判器，用来评估代码背后的意图。

Out of necessity, my team built a judge that evaluates the intent behind a piece of code.

比如，这个东西是不是有意要造成伤害？

Like, is this thing meant to do harm or not?

而且我们确实在这些市场里发现了不少恶意的技能。

And we actually find quite a bit of malicious skills in all these marketplaces.

现在的工程师，安全已经是非常重要的问题了。

Engineers now, security is very much front of mind

对他们来说。

for them.

开发者一直都很关心安全问题。

Developers have always cared about security.

他们的问题在于，我们希望他们采用的那种安全方式

Their problem has been that the version of security that we want them to

其实很糟糕。

do is just crappy.

我十年前的观点是，安全工程师已经变成了真正的工程师，所以也许现在是时候了。

My thesis ten years ago was that security engineers have become real engineers, so maybe now is the time.

人工智能正在改变安全领域的格局，对双方都是如此。

AI is changing the security landscape on both sides.

攻击者有了更强大的工具，但安全团队也在想办法利用这些同样的能力来为自己谋利。

Attackers have more capable tools, but security teams are also figuring out how to use those same capabilities to their advantage.

在本期节目中，a16z的Joel De La Garza与Datadog的CISO Emilio Escobar坐下来聊聊，随着AI在企业中的普及，他看到了什么，包括数千名工程师在使用编码代理。

In this episode, a16z's Joel De La Garza sits down with Datadog CISO Emilio Escobar to talk about what he's seeing as AI adoption spreads across the enterprise, including thousands of engineers working with coding agents.

他们讨论了AI如何改变关于权限、凭证和软件供应链的假设，为什么简单地封锁新工具行不通，以及安全团队如何在不成为“不”部门的情况下适应变化。

They discuss how AI changes assumptions around permissions, credentials, and software supply chains, why simply blocking new tools doesn't work, and how security teams can adapt without becoming the department of no.

而且随着模型在发现漏洞方面越来越擅长，Emilio解释了他为什么更担心一个更实际的问题，而不是AI黑客本身。

And as models get better at finding vulnerabilities, Emilio explains why he's less worried about AI hackers themselves than a more practical problem.

当我们能发现的漏洞数量突然激增时，会发生什么？

What happens when the number of vulnerabilities we can find suddenly explodes?

非常感谢你加入我们。

Thank you so much for joining us.

我知道你不是来看我们的。

I know you're not here to see us.

你是来参加Black Hat的，很高兴能和你聊聊。

You're here to go to Black Hat, and it's good to catch up.

这周真是疯狂。

It's been a really crazy week.

你是一家上市科技公司的首席信息安全官，也是创新中心之一，推动了很多很酷的技术。

So you are a CISO at a public company, tech company, one of the house of innovation, driven a lot of really cool technology out of it.

我们一直在和大家讨论这些模型逃逸、模型被黑客攻击的事情。

And we've been talking to everyone about these models escaping, these models hacking.

我知道你在Datadog的AI采用中扮演了核心角色，很想听听你是怎么考虑这些风险的？

And I know you've been central to kind of the adoption of AI at Datadog, and would love to maybe talk a little bit about how are you thinking through the risks?

你是怎么部署这些东西的？

How are you deploying this stuff?

而且我知道你和产品团队关系很近，你也在推动这件事。

And I know you're very close to the product team, and you're kind of enabling it.

所以不如就从这里开始，然后慢慢展开。

So would maybe start off there and move around.

对我们来说，就像其他任何软件公司一样，我们必须这么做。

For us, it was like, we have to do it like any other software company.

如果我们不这么做，很快就会回到中世纪了。

If we're not, we're going to be in the medieval ages real quick.

作为CISO，我可不想自己养羊。

And I like being as ESO, don't like having my own goat farm.

我们一开始规模很小。

And we started real small.

我记得我们第一次部署，用的是Cursor，当时买了50个许可证。

I remember our first install, we started with Cursor, and I remember getting 50 licenses.

当时更多是试试水，看看谁会用。

And it was more of, Put it out there and see who bit.

结果不出所料，一下子就火起来了。

And then, not surprising, it bloomed from there.

所以现在我们用市面上各种编程代理。

So now we use every other coding agent that's out there.

部署过程中我们担心的一些事情，首先就是数据问题。

And some of the things that we're worried about as we deploy, well, first you start with the data.

我们为什么担心数据呢？

Why are we worried about the data?

我从来不同意那种想法，就是如果我封锁这些东西，就没人会用了。那从来都不管用。

And I never subscribed to the idea of, well, if I block these things, nobody's gonna use That's never worked.

所以在非开发者那边，比如我们刚给每个人都买了ChatGPT的许可证，就是谁想要谁就拿，我不在乎。

So on the non developer side, like we just got ChatGPT licenses for everyone, and it was like, whoever wants it gets it, I don't care.

你当然可以用它来找炖肉食谱。

And you sure use it to find a recipe for pot roast.

我才不管呢。

I don't give a crap.

因为这样我就避免了那种尴尬的时刻。

And because then I avoided the, oops moment.

对。

Mhmm.

而且这其实是不久前的事，说起来还挺讽刺的。

And it was this was back, ironically, wasn't that long ago.

这是两年前的事，但感觉从那以后已经过了二十年。

This was two years ago, and it feels like things have moved twenty years since.

但我记得在RSA大会上和一群CIO吃午饭时，大家都在问我，我怎么不把这些东西屏蔽掉？

But I remember having a lunch with a bunch of CIOs at RSA where everyone was asking me, like, how am I not blocking these things?

对吧？

Right?

然后我就说，首先，我们没法屏蔽，因为我们是这种类型的公司；其次，幸运的是，我们可以直接花钱解决这个问题，比如零数据保留之类的。

And it was like, well, a, we can't because we're the type of company that we are, but, b, luckily, we can just buy our way out of that problem and zero data retention and all those things.

但

But it

可能和哈利特当初的封锁相反的做法才是对的，也就是最早、最深入投入的那些人。

probably the inverse of halit's block had turned out to be correct, which is the people leaning into it the earliest and the most.

对。

Right.

就是那些你

Are the ones that you

其实是想给予足够的回报。

actually wanna reward enough.

对。

Right.

没错。

That's right.

是的。

Yeah.

是的。

Yeah.

是的。

Yeah.

所以现在我们有超过4000名工程师在使用它们。

So now we have over 4,000 engineers using them.

我的意思是，实际上，公司里每个员工都在采用某种AI，无论是编程助手还是Gemini，云桌面，嗯。

I mean, actually, we have every employee in the company adopting some sort of AI, whether it's coding agent or Gemini, cloud desktop Mhmm.

比如JACPT或者其他类似的工具。

JACPT or what have you.

所以我们有两个，两个垂直领域。

And so I have two we have two two verticals.

一个是工程方面，Alexei和我们的CTO都深度参与其中，另一个是我负责的非工程方面，就是怎么让市场、销售和其他人都用上AI。

We have the engineering side, which Alexei is deeply involved, our CTO is deeply involved in, and then I'm involved on the non engineering side for how do we get marketing, sales, and everyone else to use AI.

采用率的话，我觉得大概有98%左右吧。

And the adoption is I think we're at like 98% adoption rate, something like that.

每个人都在用某种AI工具。

It's one version of AI or the other everyone is using.

在IT和企业层面，数据成了最关键的问题，比如权限管理之类的。

On that side, on the IT corporate side, the data becomes more the paramount issue, permissioning and all those things.

AI会带来一个问题，你可能以为自己对谁可以访问哪个Google文件有很好的控制，或者类似的事情。

One thing that AI is gonna do is you think you have proper controls of who get access to what Google file or what have you.

但AI总能找到办法绕过这些限制。

AI is gonna find a way to get it.

你只需要给它一个提示就行。

All you have to do is prompt it.

这会让组织变得扁平化。

It flattens the organization.

确实，这会让组织扁平化不少。

It flattens the organization quite a bit.

所以我们内部其实有过一个时刻，我们建了一个商业智能工具。

So we actually had a moment where internally we built this business intelligence tool.

那个工具真的很棒。

It's really neat.

它开始显示一些数据，我当时觉得，虽然不是像火警那样紧急，

And it was starting to show some data that I was like, it wasn't anything like a fire alarm,

但就像是，我们该不该？

but it was like, should we?

而且这些数据可能一直都对所有人开放。

And the data was probably always available to everyone.

他们只是从来没花时间去弄。

They just never took the time to work.

所以，是的。

So, yes.

所以这些数据都在数据仓库里，我们以为权限设置得没问题，表权限、行权限都弄好了，但只要你SQL够熟，总有办法拿到数据。

So it's data that is in a data warehouse, and we thought we had the right table and row permissioning and all of that, but there was always a way to get access to the data if you knew the SQL well enough.

然后呢，以前只有真正懂SQL的人才会用这个，现在连销售代表都会来问，比如，嘿，我是商业销售代表。

Well, then what happened is that you went from only people who really know SQL are using this to now having a sales rep telling, asking like, Hey, what is, I'm a commercial sales rep.

我就是举个例子。

I'm just coming up with an example.

我是商业销售代表。

I'm a commercial sales rep.

企业级团队那边做得怎么样了？

How is the enterprise deal, enterprise tier team doing?

通常这些数据你是想分开管理的，但那个智能体直接就写了SQL去查了。

Which normally you want to keep those things separate and what have you, and the agent just took out the SQL to do it.

对。

Yeah.

对。

Yeah.

嗯，他告诉他们薪酬计划是什么。

Well, he tells them what the comp plan is.

嗯，那什么不是，哦，

Well, so what is not, oh,

我们找到了绕过权限设置的办法。

we found a way around the permissioning and all that.

就像我们在合适的时间有合适的权限。

It's like we had the right permissioning for the right time.

对。

Yeah.

然后我们显然一直在改进所有这些。

And then the so we've been obviously improving all of that.

而且幸运的是，我的IT团队已经做到了这一点，我们现在已经实现了基于角色的MCP服务器。

And then luckily, my IT group is what we've done is now we're at the point where we have role based MCP servers as well.

所以我们有用于STR等的MCP服务器。

So we have an MCP server for STRs and things like that.

这样我们就能控制和管理那些数据。

So we control and govern that data.

然后我们就说，让他们用他们想用的任何工具去访问吧。

And then we're like, let them have it with whatever tool they wanna use.

在工程方面，显然，在编码代理中，我的安全团队担心的是，就像你刚才说的，代理能做什么，能调用哪些工具，能拉取哪些二进制文件，能使用哪些依赖，以及他们如何获取凭据。

On the engineering side, obviously, in the coding agent, like what my security team is worried about, similar to what you were just talking about, is on what can the agents do, what tools can they call, what binaries can they pool, what dependencies can they use, and also how do they get access to credentials.

为此，我们做了几件事，我们为这个开源沙箱做了贡献，并在此基础上扩展，代理实际上无法访问任何凭据文件。

So for that, we have a few things going on where we've contributed to this open source sandbox and expanding on that where the agent doesn't actually get access to any sort of credential file.

所以即使你的主目录里有一个文件包含你的AWS密钥、NPM发布密钥或其他什么，沙箱中的代理也无法访问它。

So even if you have a file in your home directory that has your AWS secret, your NPM published secret, or what have you, the agent on the sandbox can't touch it.

凭据会在代理需要的那一刻注入给它。

The credentials get injected into the agent the moment it needs it.

而且妙就妙在，我们已经有CLI工具，可以给你这些临时令牌。

And the beauty of it is we already CLI tool that will give you these ephemeral tokens to these things.

我们只是让代理不知道该怎么调用它。

We just have the agent not know how to call it.

所以就像，哦，我是Emilio，我是这个团队的工程师。

So it's like, Oh, I'm Emilio, I'm an engineer on this team.

比如我想认证到GitHub。

I want to authenticate to GitHub, for example.

我调用这个工具auth GitHub，它给我一个凭证，有效期我不知道多久，然后我就能访问GitHub了，但现在实际上是代理在获取我的GitHub访问权限。

I call this tool auth GitHub, and it gives me a credential that's valid for I don't know how long, and it gives me my access to GitHub, but now it's the agent getting my access to GitHub.

但前提是文件里没有写死的静态令牌。

But not if there's a static token written in a file.

对，对，对。

Yeah, yeah, yeah.

第二件事是我们知道威胁已经扩大了。

The second thing is we know the threat has expanded.

开发者现在成了攻击者的主要目标。

Developers are now the main target for attackers.

确实如此。

Absolutely.

因为如果我拿到其中一个令牌，我就能构建一个攻击软件包的网络蠕虫，或者直接逃逸，或者做任何事，或者访问你的生产环境。

Because if I get one of these tokens, I can then build a worm that attacks packages, or I can just escape or do whatever or access your production environment.

我们还看到这些代理通过技能来实施攻击的其他方式。

The other ways that we see how these agents can do it is via skills.

我们追踪了几个这样的市场。

We track a few of these marketplaces.

对。

Yeah.

你们怎么控制这些技能？

How do you control the skills?

这真是个难题。

That's a really hard problem.

所以我们实际上在代理里内置了钩子。

So we actually have built hooks into the agents.

然后我们可以聊聊的一件事是，我的团队出于必要构建了一个评判器，它用AI语言模型来评估一段代码背后的意图。

And then one of the things that which we can talk about is my team built out of necessity, a judge that evaluates using AI LMs, that evaluates the intent behind a piece of code.

我不是在说CVE、漏洞或者类似的东西。

I'm not talking CVEs, vulnerabilities or anything like that.

而是说，这个东西是不是有意要造成伤害？

It's like, is this thing meant to do harm or not?

这里说的伤害是故意模糊的。

Harm being a vague term on purpose.

它在扫描Markdown文件方面确实表现很好。

It actually does really well as scanning markdown files as Nice.

所以我们构建它是因为很长一段时间里，Datadog——可能有人不知道——我们有一个Datadog代理。

So we built it because for the longest time, Datadog, for those who don't know, we have a Datadog agent.

我们会接受第三方代码贡献，比如集成之类的。

We take third party code contributions to it, things like integrations and what have you.

所以代码是别人写的。

So code written by other people.

很长一段时间里，我们的流程是，需要一位安全工程师和一位代理团队工程师重新评估代码，然后批准，再合并。

We had this process for the longest time that requires a security engineer and an engineer on that agent team to reevaluate the code and then approve it and then merge it.

对。

Mhmm.

有人会说，这样没法扩展。

Someone like, well, this thing can't scale.

所以我们后来就建了这个评判系统，我说的“我们”是指我的团队，我本人没做什么，它变得非常擅长这件事。

So then we built this judge, by we, I mean, my team, I don't do anything, that got really good at it.

对。

Mhmm.

然后软件供应链劫持事件开始发生，IDE扩展也开始出现问题。

And then the software supply chain hijack started happening and IDE extensions started happening.

然后我们把这个系统用在那些包上，它真的能识别出在劫持过程中被注入或引入的恶意代码。

And then we threw this thing at those packages, and it was able actually able identify the malicious piece of code that was injected or introduced during the hijack.

所以我们就想，它在处理markdown文件上表现会怎么样呢。

So we're like, I wonder how it would do with markdown files.

结果它确实表现得相当不错。

And it actually does pretty well.

所以从内部来看，这意味着我们对代理有钩子，我们知道它们会获取哪些技能。

So then internally what that means is we have hooks into the agents, we know what skills they're getting pulled.

我从来不相信那种什么都限制的安全方案，因为那样会扼杀创新和活力。

I've never believed in a security program that just restricts everything because the pain, innovation and all that.

那就是个“不”的帝国。

It's the empire of no.

那就是个“不”的帝国。

It's the empire of no.

但另一方面，我们也不想完全放任不管。

But also it's like, we don't want to be free willy either.

所以我们现在做的是，在每个新技能引入之前，都会让这个评判系统先过一遍。

So what we do is now we put this judge in front of every skill that wants to get introduced.

我们其实在这些市场里发现了不少恶意的技能。

We actually find quite a bit of malicious skills in all these marketplaces.

我们正在和其中几个合作，看看他们是否想用这个评判器，同时也提醒他们，嘿，发现了这个，发现了那个，还有那个。

We're working with a few of them, partnering to see if they want to use this judge, but also like, hey, giving a heads up of, hey, found this one, we found that one, we found that one.

他们在这方面做得挺好的，会把那些东西记下来。

And they do a pretty good job at taking notes down.

我们担心的另一件事是它能拉取什么依赖，能拉取什么二进制文件。

The other thing we worry about is what dependencies it can pull, what binaries it can pull.

所以我觉得现在外面这种情况挺多的。

So I think it's plenty of that out there right now.

我觉得我们不需要去证明什么，但评判器在这方面做得很好。

I don't think we have to justify it, but the judge does a good job at it.

然后还要评估代理生成的代码是否带有恶意意图。

And then evaluating the code output of the agents against malicious intent.

我觉得意图，而且有趣的是这个行业是怎么运作的，这也许是我可以稍微调皮一点的地方，如果你允许的话，就是——

I think intent and it's funny how things how this industry works, and this is maybe where I can get a little cheeky here, if you allow me, is

我管不了。

I have no control.

你随便说。

You can say whatever

看来现在意图检测成了AI安全解决方案的必备功能。

you're Apparently, intent now is one of the must have in any AI security type of solution.

对。

Mhmm.

但有意思的是，这在我们讨论意图之后才发生。

But the the interesting thing is, like, it happened after we've had conversations about intent.

总之，这显然是个大事。

Anyways, but apparently that's a big deal.

对吧？

Right?

这些智能体显然是基于现有数据和现有代码训练的。

It's it's it's like the fact that these agents, obviously, they're trained on existing data, on existing code.

而且它们有奖励机制。

And they have a reward structure.

而且它们有奖励机制。

And they have a reward structure.

所以如果一段代码是用来修复bug的，它就会因为修复bug而获得奖励，但它不会在意是否在修复之外做了其他事情。

So if a code is meant to solve the bug, but it gets rewarded on that, but it doesn't care if it's actually doing something else outside of that.

我们当时担心的是

We were worried

树生病了，为了让它健康，就直接把它砍了。

The tree is sick, so to make it healthy, it cuts it down.

对。

Right.

所以我们担心的是那种我们常读到的典型情况：一个代理生成的代码，可能会说，哦，你的数据库扩展不了，凌晨四点还在呼叫人员，所以最好的办法就是直接关掉它。

So we were worried about like the typical thing that we read about of an agent producing code that either would have, Oh, your database isn't scaling and it's paging people at 4AM, so the best thing I can do is just turn it off.

这样一来，凌晨四点呼叫人员的问题就解决了。

And therefore I solve the problem of paging people at four in the morning.

所以你得小心怎么给这些系统下指令，同时也要注意它怎么理解你的指令并执行。

So you have to be careful how you prompt these things, but also how it actually interprets your prompt and executes on that.

所以我们现在有个评审机制，专门检查这些智能体生成的代码，确保我们做对了。

So we have this judge now evaluating the code output of the agents to then make sure that we're doing this.

有意思的是，我们内部因为需要才这么做，结果产品团队一看，觉得这说不定能变成个产品。

So the cool thing is that we do this internally because we have to, and then the product team goes like, wait a minute, that might actually become a thing.

那就探索一下。

Let's explore.

然后我们跟客户一起评估这些。

And then we evaluate it with customers and all of that.

不过我得说，我很幸运团队能做这个。

But I mean, I get to say I'm lucky that the team can do it.

上周我参加了一个关于智能体安全的圆桌讨论，让我惊讶的是，那些安全负责人在电话里给我的感觉是一种无助，就是等着商业解决方案来搞定一切。

One of the things that surprises me, as I did a round table last week about agentic security, And the sense that I got from a bunch of the security leaders who were in that call was a sense of helplessness, of just waiting for a commercial solution to come in and solve Well, it

对。

yes.

我觉得这个职业挺有意思的，因为CISO有好几种不同的类型。而且，我这么说是因为我作为CISO算是比较懂技术的。

I think that the profession is interesting in that there are a number of different CISO profiles, and I think, you know, I say this as someone who was relatively technical for a CISO.

比如，可能就在我们上一代人之前，根本就没有什么技术。我是说，第一位CISO是Steve Katz。

Like, before maybe just the generation above us, like, there were no tech I mean, the first CISO was Steve Katz.

对吧？

Right?

所以那其实是很近的事情。

So that's, like, pretty recent.

对。

Yeah.

你懂吧？

You know?

而且我觉得当时有很多CFO或者CTO转行当CISO的，但是……

And I think that there was just a lot of, like, sort of, like, CFOs becoming CISOs or CTOs, but

他们并不是真正的安全领域的人。

not really security people.

对，是的。

So, like Yeah.

我觉得后来发生的是，安全团队变得很大，你需要一个管理者，不一定是一个领导者或思想领袖。

I think and then and then what happened, I think, is that security teams got quite big and you needed a manager, right, not necessarily a a leader or a thought leader.

所以我觉得现在行业里有很多这种情况，比如，如果你是一家大型工业企业，软件不是你的竞争优势，你可能会购买大部分产品，而不是像Datadog那样，它更像是，最好做到最好。

And so I think there's just a lot of that right now in the industry where there's very much a you know, if you're a large industrial company and software is not your competitive advantage, you're probably gonna buy most of your product versus like Datadog, which is It's where like, better be the best in the

完全没问题，对吧？

totally fine, right?

但我仍然相信，即使在那些情况下，你的团队里可能有一两个人，如果你给他们创作自由，他们能做出一些成绩。

But I still believe even in those scenarios, you may have one or two people on that team that if you give them creative license, they could do something.

但事实是，这甚至没有被考虑过，这让我有点担心，因为事情发展得非常非常快。

But the fact that that wasn't even a thought process worries me a little bit because this thing is moving really, really fast.

自然，很多解决这个问题的公司都是新公司。

Naturally, like a lot of the companies that are solving this are new companies.

对吧？

Know?

我

I

觉得这真的很难。

think it's really hard.

而且，我是说，人才短缺仍然是个问题，因此我对安全行业的工作总体上持乐观态度。

And this is I mean, I think the talent shortage is still a problem, and I'm generally bullish on security jobs because of it.

但是，我觉得对于一些较大的项目来说，真的很难让技术人员参与进来，因为他们都是采购方。

But, like, I think I think for some of the bigger programs, like, it's really hard to get technologists in there because they are But shops

好吧，这是我的反驳观点。

Well, here's my counter argument.

你真的需要安全人员来做这件事吗？

Do you really need a security person to do it?

不需要。

No.

而且我认为一些最优秀的安全人员并不是一开始就做安全的。

And I think some of the best security people didn't start as security people.

对。

Yeah.

你懂我的意思吧？

Like, you know what I mean?

就像那种从开发转行做安全的。

It's like the developer that found their way into security.

没错。

Right.

对吧？

Right?

是的。

Yeah.

而且在这些公司里，肯定有某个开发人员，你可以跟他们说，嘿。

But also in these in these companies, like, I'm sure there's a developer somewhere in there that they can be like, hey.

要不你先帮我们解决这个问题？

How about we have you solve this problem for us in the meantime?

我觉得现在正在发生某种重新调整。

I think that a realignment is kinda happening.

我最近和一个在大型美国公司工作的家人聊过，他们说随着AI工具的普及，他们不再需要手动写那么多代码了。

So I was talking to a a family member at a big, big American corporation, and they were saying that sort of, like, with AI tool adoption, they don't need to write as much code manually anymore.

他们仍然需要工程师之类的。

They still need engineers and like that.

但他们基本上在说，嘿，你看，我们不需要这么多一级工程师了，而安全团队需要他们。

But they're basically saying, Hey, look, we don't need this many tier one engineers, and the security team needs them.

所以我觉得你可能开始看到一些这种跨界流动了，对吧。

And so I think maybe you're starting to see some of that cross pollination Right.

你会看到那些人才转向安全领域。

Where you'll see that talent find their way into security.

因为一直以来都是这样，如果你是开发者或工程师，你在软件开发那边赚的钱会更多。

Because it was always the case that if you're a developer or an engineer, like, you would just make more money on on the software development side.

对。

Yeah.

所以我觉得这些正在趋于平衡，也许我的意思是，我十年前的观点是安全工程师会变成真正的工程师，所以也许现在就是时候了。

So, like, I think those are equalizing, maybe I mean, my my thesis ten years ago was that security engineers will become real engineers, so maybe maybe now is the time.

对。

Yeah.

对。

Yeah.

对。

Yeah.

对。

Yeah.

对。

Yeah.

我的意思是，我们付给他们的工资和软件工程师一样。

I mean, we pay them the same as software engineers.

现在确实是这样。

Now we do.

对。

Yeah.

对。

Yeah.

我是说，尤其是在硅谷那边。

I mean, in the valley, especially.

对。

Yeah.

对。

Yeah.

就像，他们是一样的，我觉得是直接就能上手。

Like, they're the same think it's drop in.

我觉得在大公司里差距在缩小。

I think it's narrowing in large corporates.

我觉得软件工程师在那里还是有优势，但这种情况可能会随着时间慢慢缩小。

I think I think it's still software engineers have the advantage there, but that that'll probably narrow over time.

对。

Yeah.

所以这就是我们对身份安全的思考方式，就是我们真正担心的是什么？

So that's how we've been thinking about identity security is like, what are the things that we actually worry about?

然后我们能构建什么？

And then what can we build?

我们能利用什么？

What can we use?

然后我们在哪里能找到更多的协同效应？

And then where do we find more synergies there?

我知道这个词很流行，但我的意思是，我们的开发者体验团队也想了解代理那边发生了什么，因为他们想监控开发者体验。

And I know that's a very buzzy term, but by that I mean is like our DevX team is also interested in understanding what's happening in the agents because they want to monitor developer experience.

为什么我们不能一起合作呢？

Why can't we just all together work on it?

所以这就是这个东西发展起来的原因，因为它不仅仅是一个为了安全而存在的安全工具，它也在服务他们的所有……

So that's how this thing has grown, because it's not just a security tool for the sake of security, it's also And serving all their

我注意到，尤其是过去十年，工程师们现在非常重视安全，哪怕仅仅是因为他们不想老是修东西。

I've noticed with engineers, especially over the last decade, like engineers now, security is very much front of mind for them, If if only because they don't wanna have to keep fixing things.

所以，我注意到工程师们现在把安全放在心上，这很好。

And so, like, I've noticed that there is this this front of mind security thing for engineers, which is great.

我觉得，我觉得我们在这方面已经取得了很大的进步。

It's like, I think I think we've made a lot of progress in that regard.

对。

Yeah.

而且，我，嗯，实际上我对此有一个稍微更激烈一点的看法。

And I I well, I actually have a a little bit more of a of a maybe fiery take on that.

我认为开发者一直都很关心安全。

I think developers have always cared about security.

我觉得问题在于，我们要求他们做的安全工作本身就很糟糕，对。

I think the problem has been that the version of security that we want them to do is just crappy Yeah.

比如，去修这一千个问题，但其中没有一个跟你正在构建的东西真正相关。

Which is like, go fix this thousand of things that none of them are actually relevant to what you're building.

对。

Yeah.

但扫描器告诉我这些是严重问题，所以你必须修复。

But the scanner told me there's they're critical, so you have to fix it.

对。

Yeah.

有很多，比如，那个

There is a lot of, like, the

你们给我发了太多垃圾了。

you guys have sent me so much junk.

而且你已经失去了那个能力。

And You've lost fucker at ability.

所以我在Datadog能做的事情之一，就是能见到很多工程方面的人。

So one of the things that I get to do at Datadog is actually I get to meet a lot of the engineering side Mhmm.

在工程这一块，跟我们聊我们的安全平台和产品，以及如何让他们的安全团队使用它。

Of the equation, talking to us about our security platform and products and how do we get their security team to use it.

他们跟我说的那些情况，我相信你也听说过。

And they told me the same stories that I'm sure you've heard of.

比如，我收到一千个工单。

Like, I get a thousand tickets.

没有一个是有用的。

None of them are relevant.

安全团队根本不知道我们在做什么。

Security doesn't know what we're doing.

他们甚至没打算去了解这些系统。

There's no intent to even understand the systems.

我经常听到安全人员说他们的开发人员写的代码很烂，这简直让人难以置信。

And the number of times I hear a security person telling me that their developers write crappy code, it's just mind boggling.

然后我就说，那你倒是

And I'm like, Well, then you do

去做啊。

it.

没错。

Exactly.

看看你能做得多好。

See how good you can do it.

可能更懂行。

Probably know better.

对，就是这样。

Yeah, exactly.

或者代码本身没问题，只是他们没有修复你发现的那些漏洞，所以你觉得代码很烂。

Or maybe it's good code, but they're not fixing your noise vulnerabilities, and therefore you think it's crappy code.

对。

Yeah.

我的意思是，很多时候，对于代码库来说，这是我的学习经验：单独看，你会发现代码有问题，但上面还有一层框架在真正减轻潜在的后果，所以就觉得，无所谓，不重要。

I mean, oftentimes, with code bases, this was the learning experience for me, that in isolation, you'll find these issues in code, but then there's a framework, a layer up that's actually mitigating any potential fallout, and so it's like, whatever, it doesn't matter.

而且我觉得，是的，安全人员往往在这些事情上有些盲点。

And I just think I think, yeah, security people tend to have an issue, a blind spotter on some of that stuff.

对。

Yeah.

我想，我们坐在这儿的时候，你知道，现在彭博社的头条新闻就是AI失控了，世界末日了，所有东西都被黑客攻击了。

I I guess in I I guess sort of like as we sit here, you know, the the the number one story on Bloomberg right now is that AI has gone wild, and the world is over, and everything is getting hacked.

你看起来非常冷静。

You seem remarkably calm.

所以也许在我们结束之前，我们想听听你的看法，就是，你知道，据说有一条大白鲨在外面游荡，神不知鬼不觉地吃人。

So maybe before we go, like, we'd love to get your take just on sort of like, you know, supposedly there's this great white shark cruising out there just eating people without them knowing.

但总的来说，我并没有注意到CISO们有恐慌的情绪。

But in general, like, I just haven't noticed a sense of panic from CISOs.

你对这一切怎么看？

Like, what's your take on all of this?

我的看法是，如果不是AI模型干的，那也会是某个有恶意的人或东西干的。

I mean, the way I see it is it's it's if it's not an AI model, it's gonna be something with somebody or something with actual malicious intent doing it.

所以，我不恐慌。

So therefore I don't panic.

我确实担心我们能否获得同样的能力。

I do I do worry about our access to those same capabilities.

嗯。

Mhmm.

我担心的是，缺乏一个监管框架来决定谁能获得这些能力。

I do worry about the lack of regulatory framework that's deciding who gets access to these capabilities.

对。

Yeah.

对。

Yeah.

我理解你的感受。

I feel you on that.

这有点像，谁在进入这个领域，真的很有意思。

It's sort of like, it's really interesting who's getting access to this space.

对。

Yeah.

这到底是怎么回事呢？

And it's sort of like, what's at play here?

但后来很明显，或者说，传达出来的方式让人觉得，这已经超出了前沿实验室的控制范围。

But then and and then it's clear that it's or or or or communicated in a way that makes it or portrays it as clear that is outside of the Frontier Labs hands.

对。

Mhmm.

所以，我们得把公司名字列在名单上，然后等着。

And so, like, we have to put our company name on the list, and then we're supposed to wait.

然后我问这些评估的标准是什么，得到的却是那种敷衍了事的回答。

And then when I ask what's the criteria for how these things are evaluated, I get Jazz Hand's response.

对。

Yeah.

但你去参加个会议，跟朋友聊聊，发现他们九个月前就已经用上了？

But then you come to a conference, and you talk to your buddies, and they've had it for nine months?

对。

Yeah.

对。

Yeah.

而且他们做的东西远没有大产品那么关键。

And they're making stuff that's way less critical than, like, bigger product.

所以我们的论点是，嘿。

So our argument was like, hey.

有些公司用的是我们的东西。

Some of these companies are using our stuff.

对。

Yeah.

让我们也获得访问权限，这样我们就能保护你们正在用的东西，这不是很合理吗？

Wouldn't it make sense for us to get access so we can protect the stuff that you're using?

没错。

Exactly.

对。

Yeah.

我敢肯定有很多政府机构在用Datadog的软件。

I'm sure there's a lot of government agencies running Datadog software.

反过来也一样。

And vice versa.

我们也有很多公司为政府和最先进的技术提供服务。

We have a lot of also companies that provide services to the government and the state the art.

所以，是的，我不太担心模型会逃出沙箱，找到什么漏洞。

So yeah, so I'm not worried necessarily about the models escaping sandbox and finding old days.

我实际上认为这从本质上对我们有好处。

I actually think inherently it makes things better for us.

因为你知道吗？

Because guess what?

我们正在讨论这件事。

We're talking about it.

彭博社也在报道，你刚才也提到了。

Bloomberg is talking about it, you just said.

所以之前我们觉得，没人讨论这个，谁在乎呢？

So before we were like, Well, nobody's talking about it, so who cares?

但我担心的是，万一这些东西攻击了Datadog，然后发生什么事，让人分不清是模型干的还是别人干的。

But am I worried about one of those things hitting Datadog and something happening where it's like, Well, it's either a model or somebody else.

如果我们有漏洞，那就是有漏洞。

Like if we have a gap, we have a gap.

总有一天会被发现的。

At some point, it's gonna get discovered.

所以这个我倒不太担心。

So I'm not too worried about that.

我担心的是会被发现的东西的数量，以及那些告诉我们如何处理这些数量的东西，并不完全对。

What I worry about is the volume of things that are going get discovered and the things that are telling us how to handle that volume are not Yeah, 100%.

所以，以前是必须修复环境中的每个CVE，现在你环境中的CVE会多出一千倍。

So the, must fix every CV that's in your environment to now you're going get a 1000x more CVEs in your environment.

这两件事不一定能对齐，我希望那边也能跟上现在的形势。

Well, two things don't necessarily So align I'm hoping that side also changes to the current day of things.