Okta Oktane 2026 Conference Report

Critical Takeaways, Technologies, Trends, and Case Studies

Conference Dates: September 22-24, 2026
Location: Caesars Forum, Las Vegas, Nevada
Attendance: Over 4,000 attendees

Executive Summary

Oktane 2026 marked a pivotal moment in enterprise identity security, with Okta’s CEO Todd McKinnon announcing the Blueprint for the Secure Agentic Enterprise and forming the Blueprint Alliance. The conference demonstrated that AI agents have moved from consumer novelty to enterprise reality, requiring new governance frameworks.

Key Technologies Announced

1. Okta for AI Agents Platform

A comprehensive suite of capabilities for managing AI agent identity and security:

Shadow AI Agent Discovery for Endpoints

• Discovers unmanaged agents running on employee devices before they become blind spots
• Q3 2026 general availability

Agent Gateway

• Sits in execution path between agent and tool call
• Enforces policy and logs every interaction at runtime
• Provides deterministic kill switch to revoke active tokens instantly
• Q3 2026 general availability

Configuration Designer

• Visually maps agent-to-resource connections
• Enables governance, scoping, and auditing of handoffs
• Q4 2026 general availability

Resource Access Certifications

• Reviews agent connections over time
• Prevents standing and excessive permissions
• Generally available now

2. Agent SSO (Single Sign-On)

Built on the Cross App Access open protocol:

• Replaces non-expiring keys with short-lived, identity-governed tokens
• Reduces user consent fatigue
• Provides common way to find agents and understand their capabilities
• Generally available now

3. Agent-to-Agent Connections

• Sets rules for which agents can call other agents
• Defines what each agent can access
• Captures handoff in auditable chain
• Generally available now

4. Blueprint Alliance

Industry coalition with founding members:

• AWS, CrowdStrike, Databricks, Google Cloud, Salesforce, ServiceNow

Alliance Principles:

1. Treat every agent as a first-class identity
2. Scope access to the task rather than granting standing access
3. Keep delegation traceable
4. Monitor runtime behavior continuously
5. Enable containment that is instant and reversible

5. Okta Identity Governance (OIG) Innovations

• Advanced Entitlement Management for AWS: Fine-grained permission tracking across developers, workloads, and AI agents
• Intelligent Request Recommendations: AI-powered insights backed by request history and usage patterns
• Automated Drift Detection & Remediation: Identifies unapproved access changes in real time
• Availability: Early access Q4 2026

6. Okta Privileged Access (OPA) Innovations

• Unified Database Security: Centralizes policy enforcement across servers, SaaS accounts, and Active Directory
• Dynamic Kubernetes Protection: Eliminates hardcoded service account tokens when pods spin up
• Network Device Coverage: Extends to routers, firewalls, and switches
• Machine-Speed Workload Protection: Authorizes automated identities at machine speed
• Availability: Generally available Q1 2027

7. 48-Hour Integrations

AI-powered integration delivery reduced from 2-3 months to just 48 hours, enabling rapid unification and governance of access for people, machines, and AI agents.

Threat Intelligence Findings & Case Studies

AI-Enabled Identity Attacks (Key Statistics)

From Okta’s threat intelligence team:

• 83% of indicators Okta confirmed as malicious were not previously detected anywhere else (only 3% flagged by 4+ other sources)
• Vishing kits adapted to defeat non-phishing-resistant MFA including push notifications with number matching
• One in five proactive notifications involved phishing domains containing “passkey”
• 736 suspicious domain notifications sent in July 2026 alone
• 10,000+ suspicious domain notifications sent by October 2025

Stolen Session Tokens from AI Providers (Infostealer Analysis)

From a 7GB infostealer dump covering 5,871 infected machines in 162 countries:

Case Study: Anthropic (Claude) Incident

• Bad actor used common infostealers to hijack Claude sessions and consume usage
• One company reported nearly $1 million in fraudulent charges
• Anthropic’s response: Signed affected users out, removed saved payment methods, refunded unauthorized charges
• Impact on individual: Software architect handed a $25,000 bill for runaway cloud instance

North Korean Fake Job Interview Campaigns

Scale:

• More than 130 identities tied to about 6,500 job interviews at more than 5,000 companies
• 27% of targeted roles were outside the US, only about half in IT

Identity Verification Data (Endorsed):

• 47% of remote IT job applications carried North Korean worker patterns (up from 11% in Q3 2024)

Case Study: Qantas Breach (Mid-2025)

• Vishing attack where caller persuaded contact center agent to connect modified Salesforce Data Loader
• 5.12 million Australians affected (Qantas reported 5.7 million globally)
• Office of Australian Information Commissioner decided not to open formal investigation (as of July 2026)

AI Account Farming and Resale

Poison Claude:

• Pools accounts funded with AWS Bedrock new-account credits
• Charges 5-15% of official per-token price
• Exposed API showed 881 users

AI Video Service Abuse:

• Logged 105,000 brute-force sign-up attempts from 251 IP addresses in one month

Industry Threat Statistics (2026)

Case Studies Presented at Conference

World Central Kitchen (September 24, 2026)

• Focus: Accelerating secure AI adoption with Okta
• Outcome: Deploying AI agents for food distribution and supply chain management while maintaining enterprise security

Equals (UK Fintech) – James Simcox, Chief Product Officer

Approach: Treats agents like colleagues

• Each agent receives its own identity with tightly restricted access
• Same security controls used for employees and AI agents
• Records reasoning behind agent actions for regulatory compliance

Governance Strategy:

• Allows agents to draft email but not send it (prevents accidental mass messaging)
• Customer-facing agents can provide account settings and basic support
• Password resets and routing changes require identity validation
• Agents perform tasks directly only where Equals controls the full journey
• Payment instructions through MCP server are prohibited; access is read-only or allows writes for non-sensitive data

Large Regulated Asset Manager (Anonymized)

Agent Discovery Results:

• Scanned environment found approximately 13,000 agents
• Only about 1,000 considered valid
• More than 500 MCP servers against handful approved
• Many agents came from platforms and development environments including Microsoft-generated agents when new environments are spun up

Press Releases and Key Announcements

September 22, 2026 – Okta Platform Keynote

Title: “New Okta for AI Agents innovations increase visibility into agent behavior, secure connections at runtime, and enforce continuous agent governance”

• Announced new Okta for AI Agents capabilities
• Extended shadow AI discovery to endpoints
• Enabled visual configuration of agent connections
• Expanded Kill Switch to revoke active tokens at Agent Gateway
• Availability:
  • Generally available now: Agent SSO, Agent-to-Agent Connections, Resource Access Certifications
  • Q3 2026: Shadow AI Agent Discovery for Endpoints, Agent Gateway
  • Q4 2026: Configuration Designer, Expanded Kill Switch capabilities

September 22, 2026 – Blueprint Alliance Formation

Title: “Industry leaders form the Blueprint Alliance to advance a shared architecture for securing AI agents”

• Founding members: AWS, CrowdStrike, Databricks, Google Cloud, Salesforce, ServiceNow
• Committed to setting standards, developing integrations, and exchanging risk information
• Alliance Principles:
  1. Treat every agent as a first-class identity
  2. Scope access to the task rather than granting standing access
  3. Keep delegation traceable
  4. Monitor runtime behavior continuously
  5. Enable containment that is instant and reversible

September 22, 2026 – Okta Partner Awards 2026

Title: “Okta announces 2026 Partner Awards: Ecosystem leaders driving secure AI innovation”

• Recognizing ecosystem leaders who are helping build the secure agentic enterprise
• Full list available on official partner awards page

September 24, 2026 – World Central Kitchen Case Study

Title: “World Central Kitchen accelerates secure AI adoption with Okta”

• Demonstrating practical application of Okta for AI Agents in real-world scenario
• Food distribution and supply chain management use case

Sponsors and Partners

Official Sponsorship Information

• Sponsor Page: https://www.okta.com/oktane/sponsors/
• Conference Dates: September 22-24, 2026
• Location: Caesars Forum, Las Vegas, Nevada
• Attendance: Over 4,000 attendees

Blueprint Alliance Members (Founding)

These companies are key partners in the secure agentic enterprise initiative:

1. AWS
2. CrowdStrike
3. Databricks
4. Google Cloud
5. Salesforce
6. ServiceNow

Partner Awards 2026

Okta announced its 2026 Partner Award winners on September 22, recognizing ecosystem leaders who are helping build the secure agentic enterprise. The full list of award winners is available through Okta’s official partner awards page.

Critical Trends and Takeaways

1. AI Agents Have Crossed into Enterprise Reality

• No longer consumer novelty but established enterprise requirement
• Organizations under pressure to deploy AI swiftly without sacrificing security
• Central challenge: agents need access for value but that same access creates security risks

2. Identity is the Foundation of Agent Security

• Enterprises have long turned to identity providers for human identities and access
• Industry now applying same controls and standards to AI agents
• Okta introduced new innovations including Agent SSO and Shadow AI Agent Discovery

3. The Blueprint Approach

• Reference architecture answering four critical questions:
  • Where are my agents?
  • What can they do?
  • What are they doing?
  • How do I respond?
• Industry coalition (Blueprint Alliance) committed to setting standards and exchanging risk information
• Key insight: The secure agentic enterprise requires all industry players working together, not just single vendor solutions

4. Shadow AI is the Biggest Threat

• Users aren’t adhering to frameworks ensuring security, visibility into AI activity, and responsible usage
• Traditional perimeter defenses insufficient when autonomous agents connect directly to sensitive data
• Recommendation: Bring agents into Universal Directory as a single source of truth; known agents can be registered directly or imported from platforms like AWS Bedrock or Salesforce Agentforce

5. Runtime Security is Essential

• Most organizations running AI agents today can’t say, in real time, what those agents are actually doing
• One bad prompt can expose far more than it should
• Solution: Okta provides durable audit trail by capturing agent events in System Log and streaming directly to SIEMs; Agent Gateway sits in execution path enforcing policy and logging every interaction at runtime

6. Kill Switch Must Be Deterministic

• Even well-managed agents can be compromised
• Teams need ability to cut off rogue agent access instantly
• Evolution: Manual off switch (deactivating agent through admin console) → automated trigger when agent stops behaving like its given role
• Recommendation: Add agent-to-agent execution scenarios to cyber recovery exercises; reconstruct actions, trace downstream activity, identify affected data and systems, establish last known good state, reverse damage

7. Non-Human Identities Will Outnumber Humans by End of 2027

• BeyondID Identity Economy Report: In many enterprise environments, non-human identities will outnumber human identities before end of 2027
• Implication: Security teams must immediately audit non-human identities and extend access reviews to AI workflows
• Gap: 85% of security leaders believed they could detect identity-related breach within 24 hours, but fewer than 30% govern their non-human identities; rapid AI agent adoption in 2025-2026 made the gap worse

8. AI Makes Reconnaissance Far Easier for Attackers

• Okta threat intelligence researchers highlighted that AI makes reconnaissance and search for customer misconfigurations far easier for attackers
• Example: Vishing kits steered in real time, AI accounts farmed and resold on Telegram, North Korean job candidates using AI tools to obscure identities
• Countermeasures: Phishing-resistant authentication, tighter enrolment and recovery, security team that actually talks to HR

9. The Identity Economy Strikes Back

• Okta’s BeyondID released comprehensive research showing identity gaps fuel cybercrime
• Key finding: Adversaries treat SSO as a master key; organizations treat it as a security win (concentration risk)
• New pattern: Sequential attacks – compromise human identity, use to access service account with broader permissions, pivot to AI agent or automated workflow with system-level access
• Recommendation: Build detection for lateral movement starting at human identity and pivoting to non-human one; most SIEM rules written for human behavior patterns don’t apply to agents

10. Recovery Must Include Agent Scenarios

• Revoking credentials stops future activity but does not reverse database changes, restore deleted information, recall external communications, or unwind actions already started in other systems
• Recommendation: Add agent-to-agent execution scenarios to cyber recovery exercises; start with scenario making wrong high-impact decision, reconstruct its actions, trace downstream activity, identify affected data and systems, establish last known good state, reverse damage

Recommendations for Security Leaders (Prepared for 2027)

Immediate Actions:

1. Audit your non-human identities now – Not next quarter. Now. You cannot govern what you haven’t inventoried.
2. Extend access reviews to include AI agents and service accounts – If your access reviews only cover human identities, they’re covering less than half of your actual attack surface.
3. Stress-test your SSO trust relationships – Review every application federated into your identity platform and ask whether the trust relationship is still warranted and appropriately scoped.
4. Deploy cross-identity detection rules – Build detection for lateral movement that starts at a human identity and pivots to a non-human one; most SIEM rules were written for human behavior patterns.

Strategic Preparation:

1. Implement the Blueprint architecture – Adopt reference architecture answering: Where are my agents? What can they do? What are they doing? How do I respond?
2. Join or form an industry coalition – The secure agentic enterprise requires all industry players working together, not just single vendor solutions
3. Prepare for regulatory accountability – EU AI Act enforcement, SEC cybersecurity disclosure requirements, and emerging US federal AI governance standards will demand auditable identity governance for autonomous systems
4. Plan post-quantum migration – NIST has finalized post-quantum cryptography standards; governments have set compliance timelines; private-sector pressure will follow

Conclusion

Oktane 2026 demonstrated that AI agents have moved from consumer novelty to enterprise reality, requiring fundamentally new governance frameworks. The key insights are:

1. Identity is the foundation – Apply human identity controls to AI agents using Agent SSO and shadow discovery
2. Runtime security is essential – Monitor what agents are doing in real time with Agent Gateway
3. Industry collaboration matters – The Blueprint Alliance shows single vendors cannot secure the agentic enterprise alone
4. Non-human identities will dominate – Prepare for a world where machines outnumber humans by end of 2027
5. Recovery must include agent scenarios – Revoking credentials isn’t enough; you need to reverse damage from autonomous actions

The conference set clear direction: enterprises must build the secure agentic enterprise now, not in some distant future. This requires visibility and control without sacrificing innovation – a balance Okta’s Blueprint architecture aims to provide through industry collaboration.

Report Created: October 9, 2026
Sources: Official Okta press releases, conference coverage, threat intelligence reports, BeyondID research
Contact for additional information: Okta Newsroom (https://www.okta.com/newsroom/)