{"resourceId":"servicenow-brazil-release-ai-agent-deny-by-default-mcp-controls","versions":[{"version":"external-6dd1c844dc0e7048a4f75294e676bc5575163f8df47ba25912631666ec02f96b","resource":{"id":"servicenow-brazil-release-ai-agent-deny-by-default-mcp-controls","title":"A few AI agent updates in the ServiceNow Brazil release","organization":"ServiceNow (Community article by ServiceNow employee)","sector":"Enterprise platform; applies to all ServiceNow customers including public sector","geography":"Global","publishedAt":"2026-11-03 (posted 'yesterday' relative to 2026-11-04 retrieval)","publicationDate":"2026-11-03","eventDate":null,"sourceName":"ServiceNow Community (CEG AI CoE articles)","sourceLabel":"Vendor employee community post summarizing release notes","sourceUrl":"https://www.servicenow.com/community/ceg-ai-coe-articles/a-few-ai-agent-updates-in-the-servicenow-brazil-release/ta-p/3617284","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["developers-agents","data-security","governance-procurement"],"finding":"Brazil release notes add MCP server security controls (access rules, allowed tools), an Access Analyzer test for AI agent/workflow access, a redesigned AI Agent Studio with built-in automated evaluations, and deny-by-default access control for specified agentic AI record types on newly activated instances (previously allow).","sledRelevance":"Directly relevant to agencies and universities building Now Assist AI agents: deny-by-default and MCP tool allow-lists support least-privilege controls auditors expect. Note the default change applies to newly activated instances; existing instances may retain prior allow defaults.","evidence":"Short vendor-employee summary of release notes; no testing or independent validation. Availability depends on instance version and plugins.","architectureImplications":"MCP servers attached to agents now have tool allow-lists; agent teams must design explicit access grants under deny-by-default.","governanceImplications":"Supports least-privilege AI agent governance; agencies should verify whether upgraded (not newly activated) instances keep allow defaults and remediate.","securityPrivacyImplications":"Reduces risk of over-permissioned agents and unvetted MCP tools; Access Analyzer helps audit access. Does not address recent platform CVEs.","caveats":"Secondary vendor summary; full release notes not inspected here; default change limited to newly activated instances.","streamIds":["servicenow"],"roles":{"sales":"Lead with governance, not features: agencies worried about AI agents acting beyond authority can see concrete least-privilege controls in the Brazil release. This supports a governance-readiness conversation for existing ServiceNow public-sector customers. It is a vendor summary, so do not claim these controls close audit findings; offer an access review of existing agents instead.","engineering":"Engineers should check instance version and whether it is newly activated, since deny-by-default only applies there. Inventory existing AI agents and MCP servers, define tool allow-lists, and use Access Analyzer to test user access. Upgraded instances may still allow by default, so explicit ACL hardening is needed. Validate with the full release notes before design decisions.","delivery":"Delivery work is an agent access hardening sprint during Brazil upgrade: inventory agents, apply deny-by-default ACLs, restrict MCP tools, and run Access Analyzer tests. Interpretation: acceptance should be measured by zero agents with unreviewed allow access. Owners are the platform team with security sign-off. Risk is breaking existing agent workflows when tightening access."},"roleDetails":{"sales":{"customerProblem":"Public-sector ServiceNow customers lack confidence that AI agents and MCP tools are limited to least-privilege access.","stakeholders":["CISO","ServiceNow platform owner","Internal audit","AI governance board"],"discoveryQuestions":["How many AI agents and MCP servers are active?","Are your instances newly activated or upgraded?","When is your Brazil upgrade planned?","Who approves agent access today?"],"valueHypothesis":"Applying Brazil controls may reduce over-permissioned agent risk and ease audit evidence collection.","potentialEngagement":"A bounded AI agent access review and hardening plan tied to the Brazil upgrade.","unsupportedClaims":["Controls are independently validated","Deny-by-default applies to existing upgraded instances","These controls mitigate recent AI Platform CVEs"]},"engineering":{"fit":"Fits ServiceNow customers using AI Agent Studio or MCP servers who can upgrade to Brazil.","architectureIntegration":"Configure MCP server access rules and allowed tools; ACLs on agentic AI record types; Access Analyzer tests.","prerequisites":["Brazil release upgrade","Relevant plugins activated","Agent and MCP inventory"],"constraints":["Deny-by-default only on newly activated instances","Feature availability varies by version and plugins"],"security":"Enforce least privilege for agents and MCP tools; audit access with Access Analyzer.","proofOfValue":["Access Analyzer denies access for unauthorized test users","Each MCP server exposes only allow-listed tools","Compare pre/post ACL defaults on upgraded instance"]},"delivery":{"work":"Inventory agents and MCP servers, set deny-by-default ACLs and tool allow-lists, test access.","dependencies":["Brazil upgrade","Security team review","Agent owners"],"ownership":"ServiceNow platform team with CISO sign-off.","skillsAdoption":"Admins trained on AI Agent Studio redesign, ACLs and MCP controls.","governanceCheckpoints":["Security review of each agent's access","Change approval for ACL changes","Periodic access recertification"],"acceptanceCriteria":["Proposed: 100% of agents have reviewed explicit ACLs","Proposed: no MCP server with unrestricted tools"],"risks":["Breaking existing agent workflows","Assuming defaults changed on upgraded instances","Incomplete agent inventory"]}},"retrievedAt":"2026-11-05T04:10:30Z","enrichedAt":"2026-11-05T04:12:12Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Platform admins need skills in agent access modeling and MCP configuration.","procurementImplications":"No direct licensing change stated; upgrade planning may be needed to reach Brazil release.","operatingModelImplications":"Platform teams take on access review for each agent and MCP tool as a standing control.","sourceVerification":{"openedUrl":"https://www.servicenow.com/community/ceg-ai-coe-articles/a-few-ai-agent-updates-in-the-servicenow-brazil-release/ta-p/3617284","referenceExcerpt":"specified agentic AI record types in AI Agent Studio use deny-by-default access control. Previously, they defaulted to allow.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}