AI Security AI安全 4h ago Updated 1h ago 更新于 1小时前 43

Valarian Raises $50 Million for Sovereign Infrastructure Control Layer Valarian筹集5000万美元用于主权基础设施控制层

Valarian secured $50 million in Series A funding to expand its sovereign AI infrastructure platform. The core technology, ACRA, is a control layer built on Kubernetes that governs AI models and agents with strict isolation policies. The platform ensures data sovereignty by allowing customers to retain encryption keys and operating across cloud, on-premises, and air-gapped environments. Workloads are isolated in enclaves with default-deny networking, enabling runtime isolation of compromised agen 英国主权基础设施公司Valarian完成5000万美元A轮融资,由NEA等知名机构领投。 推出ACRA控制层,基于Kubernetes构建,用于治理组织内的AI模型、Agent及工作负载。 采用零信任架构,支持多云/本地/离线部署,客户持有加密密钥,确保数据主权。 提供运行时隔离与撤销能力,无需重构现有应用即可兼容,分为企业版和国防版。

65
Hot 热度
60
Quality 质量
60
Impact 影响力

Analysis 深度分析

TL;DR

  • Valarian secured $50 million in Series A funding to expand its sovereign AI infrastructure platform.
  • The core technology, ACRA, is a control layer built on Kubernetes that governs AI models and agents with strict isolation policies.
  • The platform ensures data sovereignty by allowing customers to retain encryption keys and operating across cloud, on-premises, and air-gapped environments.
  • Workloads are isolated in enclaves with default-deny networking, enabling runtime isolation of compromised agents without system downtime.
  • Valarian targets regulated industries and government/military sectors through distinct Enterprise and Defence deployment tracks.

Why It Matters

This development highlights the growing industry demand for secure, sovereign infrastructure for AI workloads, particularly as organizations seek to mitigate risks associated with third-party cloud providers and uncontrolled AI agents. For AI practitioners and security architects, Valarian’s approach demonstrates how to integrate governance and isolation directly into the infrastructure layer rather than relying solely on application-level fixes. The focus on air-gapped and on-premises capabilities signals a critical shift toward self-hosted, high-security AI deployments for sensitive sectors.

Technical Details

  • ACRA Control Layer: A governance layer sitting atop Kubernetes that manages identity, policy enforcement, and network segmentation for AI workloads.
  • Enclave Architecture: Each workload runs in a dedicated enclave with a specific policy stack, including short-lived secrets, scoped messaging permissions, and audit logging.
  • Runtime Isolation: The system allows for the immediate isolation, sealing, or revocation of misbehaving or compromised workloads at runtime without disrupting the broader system.
  • Cloud-Agnostic Deployment: The platform supports public cloud, on-premises, and air-gapped networks, ensuring consistent behavior regardless of the underlying infrastructure.
  • Customer-Owned Keys: Customers retain full control over encryption keys, ensuring that the vendor (Valarian) has no access to data post-deployment, thereby guaranteeing true data sovereignty.

Industry Insight

  • Shift to Sovereign AI Infrastructure: Organizations in regulated sectors must prioritize infrastructure that guarantees data residency and control, moving beyond standard SaaS AI solutions to self-hosted or hybrid models with strict governance layers.
  • Security by Design for AI Agents: As AI agents become more autonomous, embedding security controls like default-deny networking and runtime isolation at the infrastructure level is essential to prevent lateral movement in case of compromise.
  • Market Consolidation for Defense and Enterprise: The split between Enterprise and Defence tracks suggests that future AI infrastructure vendors will need to offer specialized compliance and security features tailored to the stringent requirements of government and highly regulated industries.

TL;DR

  • 英国主权基础设施公司Valarian完成5000万美元A轮融资,由NEA等知名机构领投。
  • 推出ACRA控制层,基于Kubernetes构建,用于治理组织内的AI模型、Agent及工作负载。
  • 采用零信任架构,支持多云/本地/离线部署,客户持有加密密钥,确保数据主权。
  • 提供运行时隔离与撤销能力,无需重构现有应用即可兼容,分为企业版和国防版。

为什么值得看

本文揭示了AI基础设施领域向“主权云”和“可控AI”转型的关键趋势,强调了在引入AI时数据安全与合规性的核心地位。对于关注AI安全、私有化部署及政府/金融级合规的从业者而言,提供了具体的架构参考和市场动态。

技术解析

  • ACRA控制层架构:Valarian开发了名为ACRA的控制层,直接部署在Kubernetes之上,旨在统一管理AI模型、智能体(Agents)及其他计算工作负载,解决大规模AI部署中的治理难题。
  • 细粒度安全策略:每个工作负载被置于独立的隔离区(enclave),配备专用的策略栈,涵盖身份验证、默认拒绝的网络分段、短期秘密管理、受限的消息传递权限及审计日志,实现最小权限原则。
  • 云无关性与数据主权:平台设计为云无关,可在公有云、本地数据中心或气隙网络中运行。关键特性是客户保留加密密钥,部署完成后Valarian本身无法访问数据,确保严格的数据主权。
  • 运行时弹性与兼容性:系统支持在运行时对异常或受损的工作负载进行隔离、密封或撤销,且不影响周围系统。架构设计兼容现有应用程序,无需重新构建即可集成。

行业启示

  • AI治理成为刚需:随着AI Agent和复杂模型的普及,传统的边界防御已不足够,基于Kubernetes的原生控制层和细粒度策略执行将成为企业级AI部署的标准配置。
  • 主权云市场细分加剧:针对政府和受监管行业(如金融、医疗)的“主权云”解决方案正在形成独立赛道,强调数据本地化、密钥自控和合规性,与传统公有云厂商形成差异化竞争。
  • 安全左移与架构原生:安全不再是事后附加功能,而是需要从底层架构设计开始融入(Security by Design)。企业需评估现有AI基础设施是否具备运行时隔离和数据主权保障能力。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Funding 融资 Security 安全