<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>The Permission Layer: new AI agent governance tools</title>
    <link>https://thepermissionlayer.com/launches/</link>
    <description>Products and features for governing AI agents that vendors announced in the last 90 days: agent identity, permissions, MCP security, AI data loss prevention, shadow AI discovery and spend controls. Each entry links the vendor&apos;s own announcement. Inclusion is editorial and never paid.</description>
    <language>en-us</language>
    <lastBuildDate>Wed, 18 Nov 2026 00:00:00 GMT</lastBuildDate>
    <atom:link href="https://thepermissionlayer.com/launches.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Agent 365 and Entra Suite discovery and governance of local and multicloud agents (Microsoft)</title>
      <link>https://thepermissionlayer.com/launches/#microsoft-agent-365-entra-agent-discovery-governance</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#microsoft-agent-365-entra-agent-discovery-governance</guid>
      <pubDate>Tue, 17 Nov 2026 00:00:00 GMT</pubDate>
      <category>Shadow-AI discovery</category>
      <description>Microsoft. Public preview, announced 2026-11-17. Microsoft Agent 365 and Microsoft Entra Suite discover 60 local agent types, among them Claude Code, OpenAI Codex and GitHub Copilot, plus third-party cloud agents on AWS and Google Cloud. Each discovered agent receives a Microsoft Entra Agent ID and can be governed with Conditional Access, Intune policies contain supported local agents on the device, and Dynamic Agent Risk assesses each agent&apos;s exposure, blast radius and runtime behavior. Microsoft says these capabilities are available in preview. A separate Copilot post the same day says Agent 365 is adding connected platforms including Anthropic, Databricks, OpenAI and Salesforce, without an availability date. Why it matters: Agents found on devices and in other clouds can be given an identity and placed under the same access policies as sanctioned agents, instead of only appearing in an inventory. Announcement: https://www.microsoft.com/en-us/security/blog/2026/11/17/security-for-the-agentic-era-powered-by-microsoft-perception/</description>
    </item>
    <item>
      <title>Microsoft Purview DLP for prompts to Copilot and local agents, and grounding protection for Copilot</title>
      <link>https://thepermissionlayer.com/launches/#microsoft-purview-prompt-dlp-copilot-local-agents</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#microsoft-purview-prompt-dlp-copilot-local-agents</guid>
      <pubDate>Tue, 17 Nov 2026 00:00:00 GMT</pubDate>
      <category>Data flow &amp; AI-DLP</category>
      <description>Microsoft. Announced, announced 2026-11-17. Purview data loss prevention controls extend to prompts sent to Copilot and to local agents running on Windows. A separate grounding protection for Copilot works from sensitive information types, so content can be kept out of Copilot answers without labeling every file first. Microsoft says both enter preview in January. Why it matters: Sensitive data typed into a prompt or picked up by a local agent on Windows falls under the same DLP policies as other channels. Announcement: https://www.microsoft.com/en-us/security/blog/2026/11/17/security-for-the-agentic-era-powered-by-microsoft-perception/</description>
    </item>
    <item>
      <title>Agent observability and governance for Conversational Analytics in BigQuery and Looker (Google)</title>
      <link>https://thepermissionlayer.com/launches/#google-bigquery-looker-agent-observability</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#google-bigquery-looker-agent-observability</guid>
      <pubDate>Sat, 14 Nov 2026 00:00:00 GMT</pubDate>
      <category>Agent observability</category>
      <description>Google. Generally available, announced 2026-11-14. Conversational Analytics agents in BigQuery and Looker send Cloud Monitoring metrics and Cloud Trace spans to the agent&apos;s project, and agent owners can opt in, per agent, to logging prompts and responses into a customer-owned BigQuery table. The logs inherit the dataset&apos;s IAM, encryption keys and VPC Service Controls, and logging is skipped for any turn whose data source uses row-level security, column policy tags or masking. Google lists conversation logging, the BigQuery agent monitoring dashboards (usage, sessions, latency) and the Looker System Activity dashboards (engagement and token use) as generally available. Daily input and output token quotas per project, user or agent, and native agent evaluations, are in preview. Why it matters: Teams running data agents can keep an audit trail of what each agent was asked and answered in storage they control, and can preview a cap on token use per agent. Announcement: https://cloud.google.com/blog/products/data-analytics/enterprise-ai-agent-observability-with-bigquery-and-looker</description>
    </item>
    <item>
      <title>Okta for AI Agents</title>
      <link>https://thepermissionlayer.com/launches/#okta-for-ai-agents</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#okta-for-ai-agents</guid>
      <pubDate>Thu, 05 Nov 2026 00:00:00 GMT</pubDate>
      <category>Agent identity &amp; NHI</category>
      <description>Okta. Generally available, announced 2026-11-05. Registers AI agents as identities in Okta&apos;s Universal Directory, finds unsanctioned agents, issues short-lived tokens on demand, brings agents into access certification, and can deactivate an agent with one switch and one audit trail. Okta says an agent gateway and further controls will follow in the months ahead. Why it matters: Teams already on Okta can inventory agents, review their access and shut one off in the same system they use for people. Announcement: https://www.okta.com/blog/ai/okta-for-ai-agents-general-availability/</description>
    </item>
    <item>
      <title>Microsoft Defender discovery of local AI agents and MCP servers</title>
      <link>https://thepermissionlayer.com/launches/#microsoft-defender-local-agent-discovery</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#microsoft-defender-local-agent-discovery</guid>
      <pubDate>Thu, 05 Nov 2026 00:00:00 GMT</pubDate>
      <category>Shadow-AI discovery</category>
      <description>Microsoft. Generally available, announced 2026-11-05. Microsoft Defender finds more than 50 supported AI agents running locally on Windows and macOS devices by observing their activity, and shows each agent together with the device, the account it runs under and the MCP servers configured for it. Why it matters: Coding and desktop agents that employees install on their own machines become visible to the security team in a tool many already run, which is the first step before any policy can apply to them. Announcement: https://www.microsoft.com/en-us/security/blog/2026/11/05/whats-new-in-microsoft-security-november-2026/</description>
    </item>
    <item>
      <title>Microsoft Execution Containers (MXC)</title>
      <link>https://thepermissionlayer.com/launches/#microsoft-execution-containers</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#microsoft-execution-containers</guid>
      <pubDate>Thu, 05 Nov 2026 00:00:00 GMT</pubDate>
      <category>Agent sandboxing</category>
      <description>Microsoft. Generally available, announced 2026-11-05. An operating system enforced containment layer for AI agents on Windows. Admins set least-privilege limits on the files, network access, processes and user interactions an agent can use, add stronger isolation for higher-risk cases, and can test and validate policies before enforcing them broadly. Microsoft says integration with Intune, Entra and Agent 365 is coming. Why it matters: The limits are enforced by the operating system outside the agent, so an agent running on an employee device cannot widen its own access. Announcement: https://www.microsoft.com/en-us/security/blog/2026/11/05/whats-new-in-microsoft-security-november-2026/</description>
    </item>
    <item>
      <title>Agent Identity JWT support (Google)</title>
      <link>https://thepermissionlayer.com/launches/#google-cloud-agent-identity-jwt</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#google-cloud-agent-identity-jwt</guid>
      <pubDate>Thu, 05 Nov 2026 00:00:00 GMT</pubDate>
      <category>Agent identity &amp; NHI</category>
      <description>Google. Public preview, announced 2026-11-05. Agent Identity issues agents standard OIDC tokens so they can authenticate to AWS, Azure, on-premises backends and MCP servers without static, long-lived API keys. Google announced it at its Security Summit 2026 alongside Auth Manager, which it lists as generally available. Why it matters: An agent built on Google Cloud can reach systems in other clouds with its own short-lived identity instead of a shared key stored in its configuration. Announcement: https://cloud.google.com/blog/products/identity-security/helping-you-build-and-defend-at-machine-speed-in-the-agentic-era</description>
    </item>
    <item>
      <title>Claude Enterprise inference hooks (Anthropic)</title>
      <link>https://thepermissionlayer.com/launches/#anthropic-claude-inference-hooks</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#anthropic-claude-inference-hooks</guid>
      <pubDate>Thu, 05 Nov 2026 00:00:00 GMT</pubDate>
      <category>Data flow &amp; AI-DLP</category>
      <description>Anthropic. Generally available, announced 2026-11-05. Generally available since November 5, 2026 after a beta that began in August. Claude Enterprise organizations can point Claude at their own AI security server; each governed prompt in claude.ai, Cowork and Claude Code waits for that server&apos;s allow or deny verdict before inference runs, and denials are recorded in the compliance Activity Feed. Why it matters: Buyers can apply an existing DLP or AI security policy to Claude traffic at the model call rather than at a network proxy. Announcement: https://claude.com/blog/claude-enterprise-security-updates</description>
    </item>
    <item>
      <title>CrowdStrike AI Kill Switch</title>
      <link>https://thepermissionlayer.com/launches/#crowdstrike-ai-kill-switch</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#crowdstrike-ai-kill-switch</guid>
      <pubDate>Tue, 03 Nov 2026 00:00:00 GMT</pubDate>
      <category>AI security platforms</category>
      <description>CrowdStrike. Generally available, announced 2026-11-03. Lets an authorized administrator stop a misbehaving AI agent and cut its network access. CrowdStrike says the Kill Switch is available today at its AI Gateway, with enforcement on the endpoint to follow. Why it matters: A stop control that sits outside the agent gives responders one place to halt it, though for now only for traffic that passes through the gateway. Announcement: https://www.crowdstrike.com/en-us/press-releases/crowdstrike-introduces-ai-kill-switch-for-ai-safety/</description>
    </item>
    <item>
      <title>IBM Agent Identity and HashiCorp Vault 2.2</title>
      <link>https://thepermissionlayer.com/launches/#ibm-agent-identity-vault-2-2</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#ibm-agent-identity-vault-2-2</guid>
      <pubDate>Tue, 27 Oct 2026 00:00:00 GMT</pubDate>
      <category>Agent identity &amp; NHI</category>
      <description>IBM. Generally available, announced 2026-10-27. Agent Identity records each AI agent&apos;s owner, lifecycle status and granted authority. HashiCorp Vault 2.2 issues agents short-lived, least-privilege credentials and revokes them in real time when an agent tries an action it is not authorized for. IBM says both are generally available; the watsonx Orchestrate Discover and Contain features are in tech preview. Why it matters: Teams that already run Vault can broker short-lived agent credentials and tie each agent to an owner without adding another vendor. Announcement: https://newsroom.ibm.com/2026-10-27-ibm-introduces-identity-and-control-for-ai-agents,-securing-frontier-intelligence-for-the-enterprise</description>
    </item>
    <item>
      <title>Salesforce Agent Fabric FinOps Dashboard and Model Wallet</title>
      <link>https://thepermissionlayer.com/launches/#salesforce-agent-fabric-finops-model-wallet</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#salesforce-agent-fabric-finops-model-wallet</guid>
      <pubDate>Tue, 27 Oct 2026 00:00:00 GMT</pubDate>
      <category>LLM gateways &amp; spend</category>
      <description>Salesforce. Generally available, announced 2026-10-27. Agent Fabric, the MuleSoft control plane for agents, adds a FinOps Dashboard that shows AI spend in one view and a Model Wallet that sets hard budget limits, stopping requests before costs go over budget. Salesforce says the dashboard, Model Wallet, Cost Optimization Policies and the Agent Kill Switch are generally available; Agent Evals arrive in beta in November. Why it matters: A limit that blocks the request before the money is spent is the control finance teams ask for, instead of a report after the bill arrives. Announcement: https://www.salesforce.com/news/stories/agent-fabric-catch-failures-control-ai-costs/</description>
    </item>
    <item>
      <title>Agent Gateway token management, cross-cloud enforcement and local agent governance (Google)</title>
      <link>https://thepermissionlayer.com/launches/#google-cloud-agent-gateway-governance-expansion</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#google-cloud-agent-gateway-governance-expansion</guid>
      <pubDate>Thu, 22 Oct 2026 00:00:00 GMT</pubDate>
      <category>MCP security &amp; gateways</category>
      <description>Google. Announced, announced 2026-10-22. New Agent Gateway capabilities that Google says come to preview this quarter or soon: token management that attributes token spend across agents and models and enforces budget guardrails, model routing policies, delegated user authorization through Ping Identity, SailPoint and Saviynt, Envoy proxy based policy enforcement points that extend governance to other clouds and hybrid environments, and governance for coding agents such as Claude Code and Antigravity CLI on managed devices through an OS-level sandbox and local governance services. None of these is available yet. The same post introduces Agent Compute Engine, a compute service with hardware-isolated runtime sandboxes for agents, which Google plans to offer in preview in the coming months. Why it matters: Policy set once in Google Cloud would reach agents running outside it, on other clouds and on employee laptops, and agent token spend would get a budget limit at the gateway rather than only in billing. Announcement: https://cloud.google.com/blog/topics/ai-infrastructure/introducing-agent-compute-engine-and-expanded-agent-governance</description>
    </item>
    <item>
      <title>NVIDIA OpenShell Reference Architecture</title>
      <link>https://thepermissionlayer.com/launches/#nvidia-openshell-reference-architecture</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#nvidia-openshell-reference-architecture</guid>
      <pubDate>Wed, 21 Oct 2026 00:00:00 GMT</pubDate>
      <category>Agent sandboxing</category>
      <description>NVIDIA. Generally available, announced 2026-10-21. Puts the open source OpenShell runtime and the Sentry watchdog, both part of NVIDIA&apos;s Open Agent Safety Platform in September, into one design with NVIDIA hardware: Vera Rubin systems for model inference, Vera CPUs that run agent harnesses and OpenShell, which sets a boundary around each agent, and BlueField-4 DPUs that run Sentry to monitor agents from outside their host. By NVIDIA&apos;s account, Sentry quarantines an agent that tries to leave its boundary within milliseconds. The design extends to RTX Spark, DGX Spark and DGX Station systems at the edge and runs on Linux and Kubernetes from Canonical, SUSE and Red Hat, on Windows through Microsoft Execution Containers, and on AWS, Google Cloud and Azure. NVIDIA says the reference architecture is now available. Why it matters: Enterprises running agents on NVIDIA infrastructure get a published design for where agent boundaries sit and what enforces them, and NVIDIA says security vendors including CrowdStrike, Okta, Palo Alto Networks and Zscaler are integrating support for it. Announcement: https://nvidianews.nvidia.com/news/openshell-reference-architecture</description>
    </item>
    <item>
      <title>BeyondTrust AI Agent Security</title>
      <link>https://thepermissionlayer.com/launches/#beyondtrust-ai-agent-security</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#beyondtrust-ai-agent-security</guid>
      <pubDate>Wed, 21 Oct 2026 00:00:00 GMT</pubDate>
      <category>AI security platforms</category>
      <description>BeyondTrust. Generally available, announced 2026-10-21. Finds AI agents on Windows and macOS endpoints, including frontier AI tools, internally hosted models, shadow AI and short-lived sub-agents, and separates what an agent does from what the person using it does, recording the human owner, device, agent session, skills and MCP activity. Each agent action is checked against policy at runtime before it runs; the product can steer the agent to an approved way to finish the task, or call for just-in-time privilege or human approval when the agent needs more access. BeyondTrust says it is generally available on its Pathfinder Platform for Windows and macOS endpoints. Why it matters: An agent no longer gets everything the employee is allowed to do: policy can restrict the agent on the endpoint without restricting the person. BeyondTrust&apos;s own research, released with the launch, counted about 1.28 million shell commands requested by coding agents from about 132,000 prompts over four weeks. Announcement: https://www.globenewswire.com/news-release/2026/10/21/3385607/0/en/phantom-labs-research-shows-ai-agents-are-acting-beyond-intent-beyondtrust-ai-agent-security-now-available-stops-them.html</description>
    </item>
    <item>
      <title>ThreatLocker Zero Trust AI Control</title>
      <link>https://thepermissionlayer.com/launches/#threatlocker-zero-trust-ai-control</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#threatlocker-zero-trust-ai-control</guid>
      <pubDate>Tue, 20 Oct 2026 00:00:00 GMT</pubDate>
      <category>AI security platforms</category>
      <description>ThreatLocker. Announced, announced 2026-10-20. Controls what AI agents can access, execute and connect to from outside the AI software, including when one request makes an agent act across other applications, scripts and browsers, and ships with recommended policies that teams can adjust. ThreatLocker says it works with its Allowlisting and Web Content Control products, which decide which AI applications and browser-based services are allowed at all. Why it matters: ThreatLocker customers can put agents under the same deny-by-default model they use for software, with boundaries that hold, by the vendor&apos;s account, even if an agent behaves unexpectedly. Announcement: https://www.prnewswire.com/news-releases/threatlocker-launches-zero-trust-ai-control-to-give-security-teams-command-over-ai-agents-302914016.html</description>
    </item>
    <item>
      <title>Microsoft Dynamic Permissions and Dynamic Conditional Access for agents</title>
      <link>https://thepermissionlayer.com/launches/#microsoft-dynamic-permissions-agents</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#microsoft-dynamic-permissions-agents</guid>
      <pubDate>Mon, 19 Oct 2026 00:00:00 GMT</pubDate>
      <category>Agent identity &amp; NHI</category>
      <description>Microsoft. Private preview, announced 2026-10-19. With Dynamic Permissions an agent starts with no authority and gets task-scoped, time-bound access that shrinks or ends with the task, inside an organization-wide limit, and higher-impact actions can require verified human approval. Dynamic Agent Risk scores each agent from identity, data, endpoint, cloud, network and runtime signals, and Dynamic Conditional Access in Entra responds by cutting permissions, blocking writes, narrowing reachable resources or stopping the agent. Microsoft also widened agent discovery to 60 local agent types and gave no availability dates. On 2026-11-17 Microsoft said Entra ID Governance dynamic permissions and Conditional Access dynamic controls remain in limited preview. Why it matters: Agent access follows the task and the agent&apos;s current risk instead of the standing access of the person it works for. Announcement: https://www.microsoft.com/en-us/security/blog/2026/10/19/agents-are-your-newest-insiders-security-must-move-at-their-speed/</description>
    </item>
    <item>
      <title>Evervault Browser Tokens</title>
      <link>https://thepermissionlayer.com/launches/#evervault-browser-tokens</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#evervault-browser-tokens</guid>
      <pubDate>Mon, 19 Oct 2026 00:00:00 GMT</pubDate>
      <category>Agent payments</category>
      <description>Evervault. Generally available, announced 2026-10-19. Gives a browsing agent stand-in values for a card number, CVV or password, generated in the same format as the real ones, so the agent can fill in an ordinary checkout or sign-in form. The browser&apos;s traffic runs through Evervault&apos;s proxy, which swaps in the real values only for the merchant, payment processor or gateway hosts named when the token was created; any other site receives the stand-in. Tokens expire after a set time, 15 minutes by default, their permissions cannot be widened once created, and Evervault logs each reveal and refusal per session. Evervault says Browser Tokens are available with Chromium-based headless browsers such as Browserbase and can be used today. Why it matters: The real card details never reach the agent, its logs or the model provider, and a page that tricks the agent into sending them elsewhere gets only the stand-in. Pricing: Sandbox apps are free for testing. Production apps need the pro plan, from $0.25 per card session covering the card number and CVV, and from $0.05 for non-card data such as passwords. Announcement: https://evervault.com/blog/introducing-browser-tokens-for-agentic-payments</description>
    </item>
    <item>
      <title>ServiceNow AI Passport</title>
      <link>https://thepermissionlayer.com/launches/#servicenow-ai-passport</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#servicenow-ai-passport</guid>
      <pubDate>Thu, 15 Oct 2026 00:00:00 GMT</pubDate>
      <category>Agent identity &amp; NHI</category>
      <description>ServiceNow. Announced, announced 2026-10-15. Gives each AI agent a cryptographically signed credential, grounded in ServiceNow CMDB data, GRC policies and workflow approvals, that the agent checks at runtime to confirm its authorized scope, including inside NVIDIA OpenShell. An agent that drifts can be redirected, restricted or shut down, delegated scope narrows at each step, and the credentials follow open specifications so other identity providers and enforcement points can issue and verify them. ServiceNow expects it, with shadow AI detection and AI Cost Management, in Q4 2026. Why it matters: Third-party agents get a permissions record the enterprise can verify, rather than relying on what the agent claims about itself. Announcement: https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-advances-its-operational-harness-for-enterprise-AI-with-AI-Passport-and-expanded-governance-capabilities/default.aspx</description>
    </item>
    <item>
      <title>Perplexity SPACE (open source)</title>
      <link>https://thepermissionlayer.com/launches/#perplexity-space-open-source</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#perplexity-space-open-source</guid>
      <pubDate>Thu, 15 Oct 2026 00:00:00 GMT</pubDate>
      <category>Agent sandboxing</category>
      <description>Perplexity. Generally available, announced 2026-10-15. The sandbox platform behind Perplexity Computer, now open source. Each sandbox runs in a small virtual machine with its own operating system, and network access and platform-held credentials are managed outside the sandbox, so agents reach authorized services without those credentials sitting in their workspace. In Perplexity&apos;s own testing, 108 attempts produced no escape from the virtual machine to the host; some got around network restrictions, which Perplexity says it fixed. Why it matters: Teams building their own agents can inspect and adapt a sandbox that has run every Perplexity Computer session since July, at millions of sandbox creations a day by Perplexity&apos;s count. Announcement: https://www.perplexity.ai/hub/blog/open-sourcing-space</description>
    </item>
    <item>
      <title>Descope Agentic Identity Hub 3.0</title>
      <link>https://thepermissionlayer.com/launches/#descope-agentic-identity-hub-3</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#descope-agentic-identity-hub-3</guid>
      <pubDate>Thu, 15 Oct 2026 00:00:00 GMT</pubDate>
      <category>Agent identity &amp; NHI</category>
      <description>Descope. Announced, announced 2026-10-15. Adds an approval workflow in which employees request an agent and the apps, APIs, data and MCP servers it needs, and nothing is provisioned until review, so each agent has a documented purpose, an owner and least-privilege access. It integrates with MCP gateways, including the open source agentgateway, to apply one authentication, authorization and logging layer to internal agents and to clients such as Claude, Cursor and ChatGPT, and lets agents authenticate with keys, certificates, SPIFFE or workload identity federation instead of static secrets. Why it matters: Each agent gets an owner and a review before it gets access, and Descope evaluates policy on every request, so access can be narrowed or revoked centrally without reissuing credentials. Announcement: https://www.descope.com/press-release/agentic-identity-hub-3</description>
    </item>
    <item>
      <title>Runlayer network access policies for Runlayer Agents</title>
      <link>https://thepermissionlayer.com/launches/#runlayer-agent-network-access-policies</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#runlayer-agent-network-access-policies</guid>
      <pubDate>Wed, 14 Oct 2026 00:00:00 GMT</pubDate>
      <category>Agent sandboxing</category>
      <description>Runlayer. Generally available, announced 2026-10-14. Each agent gets its own allow list or block list for outbound connections from its sandbox, whether they come from a shell command, a package install, a web fetch or the browser, and every connection is tied to the agent that opened it and logged with the run, host and outcome. Private networks, cloud metadata endpoints, raw IP addresses and plain HTTP are blocked for every agent, and an agent whose policy cannot be enforced does not run. Runlayer says the policies are available for Runlayer-hosted agents and self-hosted AWS deployments. Why it matters: Limiting where an agent can send data limits what a prompt injection can exfiltrate, even when the agent can read sensitive data. Announcement: https://www.runlayer.com/blog/zero-trust-networking-for-runlayer-agents</description>
    </item>
    <item>
      <title>American Express Intent Intelligence framework</title>
      <link>https://thepermissionlayer.com/launches/#amex-intent-intelligence</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#amex-intent-intelligence</guid>
      <pubDate>Tue, 13 Oct 2026 00:00:00 GMT</pubDate>
      <category>Agent payments</category>
      <description>American Express. Generally available, announced 2026-10-13. A Create Intent API records what a customer told an agent to buy, such as the merchant and a maximum budget, and the customer authenticates and signs off before the agent transacts. The resulting Intent ID and Proof of Intent Token scope the payment credential the agent uses at checkout in place of the card number, and serve as controls for authorization, fraud checks and disputes. Amex says the framework is available through its Agentic Commerce Experiences (ACE) Developer Kit; on October 16 it told Digital Commerce 360 that no integrations are live yet. Why it matters: The issuer gets the customer&apos;s approved limits before the agent buys, so authorization and dispute decisions can check the purchase against them. Announcement: https://www.americanexpress.com/en-us/newsroom/articles/innovation/intent-intelligence-framework-agentic-commerce.html</description>
    </item>
    <item>
      <title>Project Beacon (Goodfire and Baseten)</title>
      <link>https://thepermissionlayer.com/launches/#goodfire-baseten-project-beacon</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#goodfire-baseten-project-beacon</guid>
      <pubDate>Fri, 09 Oct 2026 00:00:00 GMT</pubDate>
      <category>AI security platforms</category>
      <description>Goodfire. Announced, announced 2026-10-09. Monitors for open models served on Baseten&apos;s inference platform that flag agent behaviors such as prompt injection, sensitive data exposure, cyber misuse and reward hacking in real time. Goodfire says Baseten customers will choose which monitors run and whether a flag is logged, sent for further review, refused or re-routed. Why it matters: Teams serving open models on Baseten can add behavior monitoring without running a second model to review every output. Announcement: https://www.goodfire.com/blog/goodfire-baseten-partnership</description>
    </item>
    <item>
      <title>Kong Token Vault</title>
      <link>https://thepermissionlayer.com/launches/#kong-token-vault</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#kong-token-vault</guid>
      <pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
      <category>Agent identity &amp; NHI</category>
      <description>Kong. Generally available, announced 2026-09-30. Agents authenticate to Kong and never hold downstream credentials: Kong Identity decides what each agent may reach, and Token Vault adds the needed credential to authorized traffic. Access can be revoked centrally, and Kong logs which identity reached which resource. Why it matters: An agent that holds no stored credentials has none to leak if it is prompt-injected or compromised. Announcement: https://konghq.com/blog/product-releases/token-vault</description>
    </item>
    <item>
      <title>Kong AI Registry</title>
      <link>https://thepermissionlayer.com/launches/#kong-ai-registry</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#kong-ai-registry</guid>
      <pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
      <category>MCP security &amp; gateways</category>
      <description>Kong. Announced, announced 2026-09-30. A catalog in which agents, MCP servers and other AI resources move through register, approve and publish steps before agents can discover them, with owner, version, approval status, risk and policies on each entry. Separate registries can show different resources to different audiences. Kong says it will be generally available at Kong AI Summit 2026. Why it matters: An approval step before discovery gives security teams a checkpoint on which MCP servers agents can find at all. Announcement: https://konghq.com/blog/product-releases/ai-registry</description>
    </item>
    <item>
      <title>Kong Advanced AI Observability</title>
      <link>https://thepermissionlayer.com/launches/#kong-advanced-ai-observability</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#kong-advanced-ai-observability</guid>
      <pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
      <category>Agent observability</category>
      <description>Kong. Announced, announced 2026-09-30. Groups the model calls, tool and MCP invocations, and policy and guardrail decisions in an agent session into one trace, with token use and cost per request and per session. Kong describes it as coming soon to Kong Konnect. Why it matters: A session-level trace shows what an agent did across a whole task, which single-request gateway logs do not. Announcement: https://konghq.com/blog/product-releases/advanced-ai-observability</description>
    </item>
    <item>
      <title>OneTrust CORIE and AI Control Plane</title>
      <link>https://thepermissionlayer.com/launches/#onetrust-corie</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#onetrust-corie</guid>
      <pubDate>Tue, 29 Sep 2026 00:00:00 GMT</pubDate>
      <category>AI governance &amp; compliance</category>
      <description>OneTrust. Announced, announced 2026-09-29. Turns an organization&apos;s policies into rules enforced at each agent tool call, allowing an action, stopping it or sending it for human review, and records the agent, request, purpose, outcome and governing policy. OneTrust&apos;s AI Control Plane applies this at runtime as a control layer between agents and enterprise systems; an MCP Gateway that lets developers build these controls into their own agents is set for private preview this fall. Why it matters: Privacy and AI governance teams can enforce policy where agents act, not only in assessments written before deployment. Announcement: https://www.onetrust.com/news/onetrust-introduces-corie-to-govern-ai-agents-at-machine-speed/</description>
    </item>
    <item>
      <title>NVIDIA Open Agent Safety Platform (OpenShell and Sentry)</title>
      <link>https://thepermissionlayer.com/launches/#nvidia-open-agent-safety-platform</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#nvidia-open-agent-safety-platform</guid>
      <pubDate>Mon, 28 Sep 2026 00:00:00 GMT</pubDate>
      <category>Agent sandboxing</category>
      <description>NVIDIA. Generally available, announced 2026-09-28. OpenShell is an open-source runtime that sandboxes agents with kernel-level file and process controls, checks outbound traffic against policy, and keeps real API keys outside the agent. Sentry is a reference design that runs a policy watchdog on BlueField-4 DPUs, apart from the agent&apos;s host, and can quarantine an agent that leaves its boundary. Why it matters: Enforcement on separate hardware, by NVIDIA&apos;s account, keeps working even if the agent&apos;s host is compromised. Announcement: https://nvidianews.nvidia.com/news/open-agent-safety-platform</description>
    </item>
    <item>
      <title>Noma agent security and governance for employee endpoints (Noma Security)</title>
      <link>https://thepermissionlayer.com/launches/#noma-endpoint-agent-security</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#noma-endpoint-agent-security</guid>
      <pubDate>Mon, 28 Sep 2026 00:00:00 GMT</pubDate>
      <category>AI security platforms</category>
      <description>Noma Security. Announced, announced 2026-09-28. Uses existing EDR or MDM data to inventory agents, MCP servers and skills on employee laptops, labels each as approved, needing review or blocked, and applies tool- and action-level policy tied to the person behind the agent. Its runtime layer can alert, block, mask data or route an action to a human. Why it matters: Coding agents on laptops come under the same policies Noma applies to SaaS and in-house agents. Announcement: https://www.prnewswire.com/news-releases/noma-extends-agent-security-and-governance-to-the-employee-endpoint-302890925.html</description>
    </item>
    <item>
      <title>Purview data security on the network with Entra Global Secure Access (Microsoft)</title>
      <link>https://thepermissionlayer.com/launches/#microsoft-purview-network-data-security-ai</link>
      <guid isPermaLink="true">https://thepermissionlayer.com/launches/#microsoft-purview-network-data-security-ai</guid>
      <pubDate>Thu, 24 Sep 2026 00:00:00 GMT</pubDate>
      <category>Data flow &amp; AI-DLP</category>
      <description>Microsoft. Generally available, announced 2026-09-24. Entra Global Secure Access enforces Purview classification and data policies at the network layer, for people and for agent traffic acting on a user&apos;s behalf, and can block sensitive files or text from reaching risky destinations such as unsanctioned AI tools. Why it matters: Agent traffic falls under the same data policies as employee uploads, without a separate AI-specific DLP product. Announcement: https://www.microsoft.com/en-us/security/blog/2026/09/24/whats-new-in-microsoft-security-september-2026/</description>
    </item>
  </channel>
</rss>
