Daily enterprise tech brief ·
Enterprise AI releases: 28 September 2026
What changed. Why it matters to your business. What to test.
Lead signals
Research Hub ↗Collibra advances runtime agent governance.
What's new: Collibra made Maestro available, put Studio and Assistant in preview, and plans Guardian Agents runtime supervision through AI Command Center in October.
Governance teams can define permissions, tool access, certified sources, and machine-readable behavioral contracts before runtime controls flag or block unauthorized agent actions.
Oracle adds passwordless OCI passkeys.
What's new: OCI IAM now supports FIDO2 discoverable credentials, enabling a Continue with Passkey sign-in flow without usernames or passwords.
Security teams can require passkeys for high-risk groups, reducing phishing-prone password use while retaining policy-based enrollment and authentication controls.
Oracle adds long-term agent memory.
What's new: Oracle added project-scoped long-term memory to Enterprise AI Agents, selectively storing embeddings and retrieving context under conversation-level recall and storage policies.
Teams can reduce repetitive context collection for persistent agents while explicitly controlling when conversation-derived information is recalled, stored, or excluded.
Checkmarx brings API risks into MCP.
What's new: Checkmarx MCP Server added generally available API Security support for Claude Code, Cursor, Windsurf, and Copilot CLI, plus expanded IaC and Secrets triage support.
AppSec teams can bring exposed and shadow API context into developer AI clients, reducing context switching when investigating and initiating remediation.
Zendesk expands AI knowledge grounding.
What's new: Zendesk made dedicated Helpjuice, WordPress.com, and ReadMe connectors generally available for synchronizing external knowledge into Zendesk Knowledge.
Support teams can ground AI agents, Copilot, search, and generative search in centrally synchronized documentation instead of manually duplicating content.
HubSpot launches Claude sales workflows.
What's new: HubSpot released a free Sales Plugin for Claude with eight CRM skills and user confirmation before proposed record changes are written.
Sales teams can prepare calls, review pipelines, log activities, and draft follow-ups conversationally, while confirmation gates reduce unintended CRM updates.
Oracle enables governed agent reuse.
What's new: Oracle enabled agent-team sharing across database schemas through owner-qualified teams and grants to named users or roles.
Platform teams can reuse governed agent workflows without copying definitions, while resource privileges for tools, credentials, tables, and functions remain separately enforced.
Executive summary & decision signals
POC focus: Governance teams can define permissions, tool access, certified sources, and machine-readable behavioral contracts before runtime controls flag or block unauthorized agent actions.
- Governance teams can define permissions, tool access, certified sources, and machine-readable behavioral contracts before runtime controls flag or block unauthorized agent actions.
- Security teams can require passkeys for high-risk groups, reducing phishing-prone password use while retaining policy-based enrollment and authentication controls.
- Teams can reduce repetitive context collection for persistent agents while explicitly controlling when conversation-derived information is recalled, stored, or excluded.
All source evidence
Practical questions
What enterprise change is described in "Collibra advances runtime agent governance."?
Collibra made Maestro available, put Studio and Assistant in preview, and plans Guardian Agents runtime supervision through AI Command Center in October.
What enterprise change is described in "Oracle adds passwordless OCI passkeys."?
OCI IAM now supports FIDO2 discoverable credentials, enabling a Continue with Passkey sign-in flow without usernames or passwords.
What enterprise change is described in "Oracle adds long-term agent memory."?
Oracle added project-scoped long-term memory to Enterprise AI Agents, selectively storing embeddings and retrieving context under conversation-level recall and storage policies.
What enterprise change is described in "Checkmarx brings API risks into MCP."?
Checkmarx MCP Server added generally available API Security support for Claude Code, Cursor, Windsurf, and Copilot CLI, plus expanded IaC and Secrets triage support.
What enterprise change is described in "Zendesk expands AI knowledge grounding."?
Zendesk made dedicated Helpjuice, WordPress.com, and ReadMe connectors generally available for synchronizing external knowledge into Zendesk Knowledge.
What enterprise change is described in "HubSpot launches Claude sales workflows."?
HubSpot released a free Sales Plugin for Claude with eight CRM skills and user confirmation before proposed record changes are written.
What enterprise change is described in "Oracle enables governed agent reuse."?
Oracle enabled agent-team sharing across database schemas through owner-qualified teams and grants to named users or roles.
Business implications are editorial assessments, not measured returns. Confirm release status, availability and eligibility before testing.
