Daily enterprise tech brief ·
Enterprise AI releases: 25 September 2026
What changed. Why it matters to your business. What to test.
Lead signals
Research Hub ↗BNP Paribas Commits to Governed Banking Agents
What's new: BNP Paribas signed a five-year Google Cloud agreement for Gemini Enterprise, Gemini models, AI infrastructure, and governed banking-agent use cases.
The model uses authenticated, least-privilege agents and monitored interactions to support controlled credit and research workflows.
Gemini Live Avatars Reach General Availability
What's new: Google Cloud made Gemini 3.8 Live with Live Avatar generally available in Gemini Enterprise, supporting speech, video, tools, vision, and 97 languages.
Teams can build multilingual service agents that maintain conversations while CRM, ERP, or API actions execute in the background.
IBM Adds Swift Tokenized-Deposit Connectivity
What's new: IBM added beta Swift ISO 20022 messaging support and an on-premises IBM Z and LinuxONE deployment option to Digital Asset Haven.
Regulated institutions can evaluate tokenized-deposit workflows using familiar payment standards while retaining client-controlled infrastructure and key management.
AlloyDB Isolates Agent Queries From Production
What's new: Google Cloud previewed AlloyDB PostgreSQL for agents, creating sandboxed read-only instances for bursty agent queries against current production data.
Isolated read workloads can protect transactional systems while giving agents low-latency operational context and reducing idle compute through scale-down behavior.
Adobe AEM Adds AI Asset Controls
What's new: Adobe released AEM Cloud Service 2026.9.0 with AI asset search, Smart Collections, protected video delivery controls, and agentic authoring capabilities.
Marketing teams can find governed assets faster, apply media access rules, and validate platform changes before wider production traffic cutover.
Gemini Enterprise Adds Agent Control Layers
What's new: Google Cloud introduced Gemini Enterprise Projects, Skills Registry, and Agent Sandbox, while making Auth Manager generally available and JWT support publicly previewable.
Persistent context, centrally approved skills, isolated execution, and managed credentials can standardize controls for autonomous workflows.
Google Open-Sources Guardrailed GKE Migration
What's new: Google Cloud released an open-source agent plugin that translates EKS infrastructure and Kubernetes manifests into GKE landing zones through pull requests.
Platform teams can reduce configuration translation work while preserving GitOps review, deterministic checks, and human approval before changes apply.
Also on today's Radar
Open any update for the use case, evidence and POC request.
08GitHub Adds Step-Up Checks for Admin Actions
What's new: GitHub introduced public-preview proof-of-presence checks for sensitive enterprise actions, requiring interactive re-authentication or MFA challenges before protected changes.
Action-level step-up authentication can prevent compromised sessions from creating tokens or changing high-impact security and webhook settings.
09DigiCert Operationalizes Post-Quantum Migration Planning
What's new: DigiCert made Quantum Central generally available, centralizing cryptographic data, policy application, remediation workflows, ownership assignments, and readiness tracking.
Security teams can turn cryptographic discovery into auditable, owned remediation programs across priority applications and legacy infrastructure.
10Abnormal Makes AI Governance Generally Available
What's new: Abnormal AI made AI Governance generally available to existing customers, providing AI-use discovery, policy evaluation, monitoring, spending visibility, and audit evidence.
Security leaders can identify sanctioned and unsanctioned AI use, document policy posture, and prepare defensible evidence for audit reviews.
11Iterable Connects Salesforce Data to RCS Journeys
What's new: Iterable's Fall 2026 release adds native Salesforce data synchronization, RCS messaging, catalog segmentation, Nova Agent recommendations, and webhook response caching.
Customer teams can orchestrate richer mobile journeys from synchronized CRM context while reducing custom data integration and repeated webhook load.
12Gurucul Connects AI Risk to Identity Context
What's new: Gurucul made AI Risk and Response generally available, correlating AI activity with identity, endpoint, cloud, proxy, and operating-system telemetry.
SOC teams can investigate shadow AI, risky agents, and data exposure using the identities, permissions, and systems behind each alert.
13LatticeFlow Launches Managed AI Risk Center
What's new: LatticeFlow AI launched AI Risk Center, a managed service delivering recurring technical assessments, system-level evidence, risk metrics, and review sessions.
Organizations without dedicated AI assurance teams can establish continuous technical reviews and evidence for deployment decisions and governance thresholds.
14Gurucul Monitors AI Agents Across Security Signals
What's new: Gurucul AI Risk and Response now ingests named AI-platform activity alongside endpoint, identity, cloud, proxy, and operating-system telemetry for connected investigations.
Security teams can prioritize AI incidents through linked access, data, behavior, and permission context rather than isolated usage alerts.
15Avalara Releases Agentic Tax Integration Fabric
What's new: Avalara made Versori available through its Developer Portal, using agents to plan, connect, build, test, validate, and publish tax-service connectors.
ERP, billing, and commerce teams can shorten compliance-integration work by automating schema mapping, authentication, transformation logic, and connector validation.
Executive summary & decision signals
POC focus: The model uses authenticated, least-privilege agents and monitored interactions to support controlled credit and research workflows.
- The model uses authenticated, least-privilege agents and monitored interactions to support controlled credit and research workflows.
- Teams can build multilingual service agents that maintain conversations while CRM, ERP, or API actions execute in the background.
- Regulated institutions can evaluate tokenized-deposit workflows using familiar payment standards while retaining client-controlled infrastructure and key management.
All source evidence
Practical questions
What enterprise change is described in "BNP Paribas Commits to Governed Banking Agents"?
BNP Paribas signed a five-year Google Cloud agreement for Gemini Enterprise, Gemini models, AI infrastructure, and governed banking-agent use cases.
What enterprise change is described in "Gemini Live Avatars Reach General Availability"?
Google Cloud made Gemini 3.8 Live with Live Avatar generally available in Gemini Enterprise, supporting speech, video, tools, vision, and 97 languages.
What enterprise change is described in "IBM Adds Swift Tokenized-Deposit Connectivity"?
IBM added beta Swift ISO 20022 messaging support and an on-premises IBM Z and LinuxONE deployment option to Digital Asset Haven.
What enterprise change is described in "AlloyDB Isolates Agent Queries From Production"?
Google Cloud previewed AlloyDB PostgreSQL for agents, creating sandboxed read-only instances for bursty agent queries against current production data.
What enterprise change is described in "Adobe AEM Adds AI Asset Controls"?
Adobe released AEM Cloud Service 2026.9.0 with AI asset search, Smart Collections, protected video delivery controls, and agentic authoring capabilities.
What enterprise change is described in "Gemini Enterprise Adds Agent Control Layers"?
Google Cloud introduced Gemini Enterprise Projects, Skills Registry, and Agent Sandbox, while making Auth Manager generally available and JWT support publicly previewable.
What enterprise change is described in "Google Open-Sources Guardrailed GKE Migration"?
Google Cloud released an open-source agent plugin that translates EKS infrastructure and Kubernetes manifests into GKE landing zones through pull requests.
What enterprise change is described in "GitHub Adds Step-Up Checks for Admin Actions"?
GitHub introduced public-preview proof-of-presence checks for sensitive enterprise actions, requiring interactive re-authentication or MFA challenges before protected changes.
What enterprise change is described in "DigiCert Operationalizes Post-Quantum Migration Planning"?
DigiCert made Quantum Central generally available, centralizing cryptographic data, policy application, remediation workflows, ownership assignments, and readiness tracking.
What enterprise change is described in "Abnormal Makes AI Governance Generally Available"?
Abnormal AI made AI Governance generally available to existing customers, providing AI-use discovery, policy evaluation, monitoring, spending visibility, and audit evidence.
What enterprise change is described in "Iterable Connects Salesforce Data to RCS Journeys"?
Iterable's Fall 2026 release adds native Salesforce data synchronization, RCS messaging, catalog segmentation, Nova Agent recommendations, and webhook response caching.
What enterprise change is described in "Gurucul Connects AI Risk to Identity Context"?
Gurucul made AI Risk and Response generally available, correlating AI activity with identity, endpoint, cloud, proxy, and operating-system telemetry.
What enterprise change is described in "LatticeFlow Launches Managed AI Risk Center"?
LatticeFlow AI launched AI Risk Center, a managed service delivering recurring technical assessments, system-level evidence, risk metrics, and review sessions.
What enterprise change is described in "Gurucul Monitors AI Agents Across Security Signals"?
Gurucul AI Risk and Response now ingests named AI-platform activity alongside endpoint, identity, cloud, proxy, and operating-system telemetry for connected investigations.
What enterprise change is described in "Avalara Releases Agentic Tax Integration Fabric"?
Avalara made Versori available through its Developer Portal, using agents to plan, connect, build, test, validate, and publish tax-service connectors.
Business implications are editorial assessments, not measured returns. Confirm release status, availability and eligibility before testing.
