供资
以6亿美元的价格进行估价，以确保AI Agent访问
在 Google 上的首选来源中添加 Unite.AI
Ificial公司以6亿美元的估值为B系列提供了4 000万美元的融资资金，为企业AI采用的同时出现的安全问题带来了新的资本：如何让软件agents在关键系统内工作，而不给他们解锁这些系统的资质。
该回合由Erad Gil领导，他还投资了公司的A系列。 其他投资者包括Gradient和Y Combinator. 虚构计划利用资金加快产品开发，扩大工程和上市团队，为AI agents建设安全基础设施。
旧金山公司已经提供了管理秘密，证书，加密密钥，以及特权准入的基础设施。 它的Agent Vault将安全焦点应用于需要与API和企业服务互动的agents，同时将基础密码和API密钥保留在agent的可及范围之外。
将 Agent 的访问权限从证书中分离
Agents调查某一事件、审查代码或自动操作工作流程可能需要访问寄存器、云服务和商业应用程序。 将资质直接放置在他们的环境中，为敏感信息创造了另一个途径，可以到达日志，模型提供者，或者操纵agent通过提示词注入的攻击者。
Infisical的Agent Vault，目前标注为在其产品页面上的预览，通过前置代理访问的经纪人。 agent向批准服务发送请求，代理在出道时应用实际证明。 根据服务配置，它可以附加或替换授权标题，或者替换一个占位符的证书。 称公司的基本秘密远离agent的环境、文件系统和模型背景。
实际上，这改变了信任所在之处。 agent可以要求授权的服务在不读取用于认证请求的API密钥的情况下进行工作。 与符合标准HTTPS PROXY配置的HTTP客户端兼容的不规则文档，允许该方法适应现有的agent运行时间。
agents不应该有任何不同。 称 Vlad Matsiiako, “Agents需要进入公司系统才能完成任务，但给他们实际的密码和API密钥是一个重大的安全风险。 我们让agents安全地完成工作，而从未掌握这些证书。”
定义每个 Agent 能做什么
Inficial的文档将访问捆绑描述为一个会话中agent可以使用的服务列表，包括API和MCP服务器。 会话提供有时限的访问，同时代理截取出行请求并附上真实的证书。 会话日志提供了请求的加密记录。
这种分离是有用的，但仅凭证书隔离并不能保证agent的决定安全。 agent如果允许修改一个系统，仍然可以使用该允许的访问进行不想要的修改。 因此，团队需要界定某项工作实际需要的服务和行动，以及证书控制。 相关的问题既在于agent能否获得一个秘密，也在于它能实现什么而从未见到一个秘密。
在AI agents的新身份和证书控制中已经可以看到更广泛的企业转移。 随着agents跨业务系统开展工作，安全小组需要一种方法，将授权与可观察到的行动联系起来，并在授权结束时取消访问。
在现有担保业务的基础上再接再厉
它为1300多客户和30万开发者服务，每天保守100亿秘密，为其核心开源项目赚得超过3万GitHub星。 公告中命名的客户包括空中客车，大众汽车，Hugging Face,OpenRouter,Lucid，美国第一金融公司Wise，和Clay.
这些数字描述了公司的更广阔的平台，而不是披露的客户对Agent Vault的盘点。 这一区别很重要：融资既支持既定的秘密管理业务，又支持将其扩展为agent接入基础设施。
Matsiiako 称 Infisical在融资公告中将继续投资于秘密管理，证书管理和特权访问管理，同时加大对Agent Vault的关注。 他还在称公司雇用100多个职位，包括工程、销售、营销、产品、设计和其他团队。
agents正在成为每个企业内部的新用户，“瓦莱德和联合创始人迈杜尔和托尼早早认识到AI agents会制造一系列全新的安全挑战，他们已经把非利士建成一个我认为可以成为AI时代必不可少的基础设施的平台。”
系列B为寻找这一机会提供了更多的资源。 其核心主张是具体的：企业应当能够将工作委托给agents，同时将资信置于为管理这些业务而建设的安全基础设施的控制之下。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Funding
Infisical Raises $40 Million Series B at $600 Million Valuation to Secure AI Agent Access
Add Unite.AI to your preferred sources on Google
Infisical has raised $40 million in Series B funding at a $600 million valuation, bringing fresh capital to a security problem emerging alongside enterprise AI adoption: how to let software agents work inside critical systems without giving them the credentials that unlock those systems.
The round was led by Elad Gil, who also invested in the company’s Series A. Other investors include Gradient and Y Combinator. Infisical plans to use the funding to accelerate product development, expand its engineering and go-to-market teams, and build out security infrastructure for AI agents.
The San Francisco company already provides infrastructure for managing secrets, certificates, encryption keys, and privileged access. Its Agent Vault applies that security focus to agents that need to interact with APIs and enterprise services while keeping the underlying passwords and API keys outside the agent’s reach.
Separating an Agent’s Access From Its Credentials
Agents that investigate an incident, review code, or automate an operational workflow may need access to repositories, cloud services, and business applications. Placing credentials directly in their environment creates another route for sensitive information to reach logs, model providers, or an attacker who manipulates the agent through prompt injection.
Infisical’s Agent Vault, currently labeled as a preview on its product page, brokers that access through a forward proxy. The agent sends a request toward an approved service, and the proxy applies the actual credential on the way out. Depending on the service configuration, it can attach or replace an authorization header or substitute a credential for a placeholder. The company says the underlying secret stays out of the agent’s environment, filesystem, and model context.
In practical terms, this changes where trust sits. An agent can ask an authorized service to perform work without reading the API key used to authenticate the request. Infisical documents compatibility with HTTP clients that honor the standard HTTPS_PROXY configuration, allowing the approach to fit into existing agent runtimes.
“You wouldn’t give an employee unrestricted access to every system in your company, and AI agents shouldn’t be any different,” said Vlad Matsiiako, co-founder and CEO of Infisical. “Agents need access to company systems to do their jobs, but giving them actual passwords and API keys is a major security risk. We let agents get their job done safely without ever having those credentials in their hands.”
Defining What Each Agent Can Do
Infisical’s documentation describes access bundles as lists of services an agent can use during a session, including APIs and MCP servers. Sessions provide time-bound access, while proxies intercept outbound requests and attach the real credentials. Session logs provide an encrypted record of requests.
That separation is useful, but credential isolation alone does not make an agent’s decisions safe. An agent with permission to modify a system can still make an unwanted modification using that permitted access. Teams therefore need to define the services and actions a job actually requires, alongside the credential controls. The relevant question becomes both whether an agent can obtain a secret and what it can accomplish without ever seeing one.
The broader enterprise shift is already visible in new identity and credential controls for AI agents. As agents take on work across business systems, security teams need a way to connect delegated authority with observable actions and revoke access when that authority ends.
Building on an Existing Security Business
Infisical says it serves more than 1,300 customers and 300,000 developers, secures 10 billion secrets daily, and has earned more than 30,000 GitHub stars for its core open-source project. Customers named in the announcement include Airbus, Volkswagen, Hugging Face, OpenRouter, Lucid, First American Financial Corporation, Wise, and Clay.
These figures describe the company’s wider platform, rather than a disclosed customer count for Agent Vault. That distinction matters: the financing supports both an established secrets-management business and its expansion into agent access infrastructure.
In his funding announcement, Matsiiako said Infisical would continue investing in secrets management, certificate management, and privileged access management while increasing its focus on Agent Vault. He also said the company is hiring for more than 100 positions across engineering, sales, marketing, product, design, and other teams.
“AI agents are becoming a new class of users inside every enterprise, and securing their access will be one of the biggest cybersecurity challenges of the next decade,” said Elad Gil. “Vlad and co-founders Maidul and Tony recognized early that AI agents would create an entirely new set of security challenges, and they’ve built Infisical into a platform that I believe can become essential infrastructure for the AI era.”
The Series B gives Infisical more resources to pursue that opportunity. Its central proposition is concrete: enterprises should be able to delegate work to agents while keeping credentials under the control of security infrastructure built to manage them.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.