Google发布Beyond Zero：面向AI时代的零信任安全模型

InfoQ Homepage News Beyond Zero: Google Publishes Successor to BeyondCorp Development Beyond Zero: Google Publishes Successor to BeyondCorp Sep 05, 2026 2 min read by Renato Losio Follow us on Youtube232K Followers Linkedin26K Followers InstagramNew RSS19K Readers X57.1k Followers Facebook21K Likes BlueskyNew In a recent research paper, Google introduced Beyond Zero, a "security model for the AI era" that extends Zero Trust to autonomous AI agents. The new approach moves access decisions from the application level to individual resources and actions, combining static authorization controls with dynamic AI-driven decisions to enable machine-speed enforcement for humans and agents. Beyond Zero uses contextual and risk-based, resource-level controls to continuously authorize individual actions by both humans and AI agents. The new model is based on five principles: authorization at the level of individual actions and resources across interfaces and APIs; a combination of static policies and dynamic controls for higher-risk scenarios; automatically enriched context about users, actions, data, and risks; automated investigation triggered by risk signals; and challenges or containment measures that can require additional verification or telemetry from users and AI agents. Joseph Valente, formerly director of product management at Google, and Michal Zalewski, distinguished security researcher and formerly at Google, write in the paper: The assumptions underpinning BeyondCorp — that accessors are human, that actions occur at human speed, and that applications are the correct boundary for trust—are no longer sufficient. In 2014, Google published the BeyondCorp whitepaper, outlining its approach to replacing network-perimeter security with a zero-trust model for enterprise access. To adopt the latest model, SaaS vendors will need to expose action-level authorization, standards need to mature, and smaller security teams face difficult issues around false positives, intent, auditing, and cost. On LinkedIn, Valente explains: Beyond Zero is one of those areas where Google had built something internally that was well ahead of anyone else for a long time. Doing continuous authorization of every action at scale seemed like overkill at first. But then thousands of workers became hundreds of thousands of workers. Hundreds of thousands of workers were joined by millions of agents. And so on... BeyondCorp vs. Beyond Zero. Source: ACM digital library. Describing the "new paradigm for enterprise security," Heather Adkins, VP of security engineering at Google, and Archana Ramamoorthy, senior director at Google, explain: The enterprise landscape is now entering a new era shaped by artificial intelligence, where AI is changing the assumptions around how enterprise security works. AI agents are being deployed globally to increase operational velocity and boost productivity. In the article "Beyond Zero: For The Rest Of Us," Kane Narraway, security manager at Canva, is positive about Beyond Zero's goals, but skeptical about how quickly ordinary enterprises can implement them. Narraway warns: It’s worth being clear about what this paper is and isn’t though. It’s aspirational. Google says deployments are internal-only, the components aren’t all built, and more papers are coming. This is the same playbook as 2014: publish the vision, describe the architecture at a high level, and encourage the industry to build towards it collaboratively (...) You aren’t Google and don’t try to be. The community reaction on Hacker News has been so far largely skeptical, with commenters questioning the reliability and complexity of AI-driven authorization. User stogot comments: Zero trust is deterministic. AI is non-deterministic. Non-deterministic access controls are a terrible idea. Some practitioners argue that access control is traditionally a hard security boundary, whereas probabilistic decisions are inherently harder to predict, test, and audit. Firas Durri writes: Preparing for AI agents means expanding auditability and reversibility in software. Google presents Beyond Zero as an ongoing effort and says it plans to publish further details on its implementation and operational considerations, but no specific deadlines have been published. About the Author Renato Losio Show moreShow less This content is in the Cloud Security topic Related Topics: Development Architecture & Design DevOps AI, ML & Data Engineering Enterprise Architecture Application Security AI Security Google Cloud Cloud Security Related Editorial Related Sponsors Popular across InfoQ Kubernetes Promotes KYAML as a Safer, More Consistent Way to Work with Manifests Airbnb Cuts Authentication Code by 60% with Server Driven Architecture OpenClaw 2.0 Releases with Simplified Setup and Collaborative Agents Google Mantis: an Agentic Vulnerability Scanning Harness for Reducing False Positives Java News Roundup: TornadoVM 6, JReleaser, LangChain4j, Java Operator SDK, JHipster, Yupiik Fusion Personality Over Skillset: How Adam Wachtel Builds Engineering Teams The InfoQ Newsletter A round-up of last week’s content on InfoQ sent out every Tuesday. Join a community of over 250,000 senior developers. View an example We protect your privacy.