The news
Okta (OKTA) on September 22 used its Oktane conference to broaden Okta AI agent security, announcing new controls for finding, connecting and shutting down AI agents, some available now and others planned for later quarters, and a 12-company alliance, including AWS and Salesforce, to standardize agent security.
The additions extend Okta for AI Agents, its product line for managing software agents as identities. Agent SSO swaps non-expiring keys for short-lived tokens governed by identity policy. Agent-to-Agent Connections controls which agents may call other agents and keeps an auditable record of those handoffs. Resource Access Certifications let administrators review agent permissions over time.
Okta said those three features are generally available now. Agent Gateway, which enforces policy and logs interactions between agents and tools in real time, and Shadow AI Agent Discovery for Endpoints, which finds unmanaged agents on employee devices, are planned to be generally available in Q3. Configuration Designer, a visual map of which agents connect to which resources, and an expanded Kill Switch that builds on Okta's existing manual off switch by revoking active tokens at the gateway, are planned for Q4. Okta did not say whether those are calendar or fiscal quarters, and cautioned that unreleased features may not ship on time or at all.
Okta also announced the Blueprint Alliance, whose founding members are AWS, CrowdStrike, Databricks, Docker, Google Cloud, Lovable, Okta, Proofpoint, Salesforce, ServiceNow, Wiz and Zscaler, with GE Appliances and World Central Kitchen as strategic advisers. The group has released a multi-vendor reference architecture for securing agents, built on the blueprint Okta introduced in March 2026, and plans cross-vendor signal sharing built on open standards including MCP, OCSF, SSF and CAEP, and regular joint interoperability results. Andrew Comstock, SVP and general manager of Salesforce's MuleSoft, said in the announcement that “security has to be a shared architecture, not a single vendor's feature.”
At an investor session in Las Vegas on September 23, chief executive Todd McKinnon argued that agentic security could become a bigger market than all of identity and access management and said market share would probably be set within about three years, according to a transcript summary published by Investing.com. McKinnon said that in a few sizable deals, the AI agent portion added about 25% to 37.5% over base platform pricing, but cautioned that it was a small sample. Executives at the session named Microsoft, IBM and ServiceNow as competitors.
The numbers
- Blueprint Alliance founding members
- 12
- Agents in use per average global Fortune 500 enterprise by 2028 (Gartner, cited by Okta)
- More than 150,000
- Organizations confident in agent governance (cited by Okta)
- 13%
- Pricing uplift in a few sizable AI agent deals (McKinnon, small sample)
- About 25% to 37.5%
Why CEOs should care
CISOs should start with inventory. Okta's announcement cites a Gartner prediction that an average global Fortune 500 enterprise will have more than 150,000 agents in use by 2028, and a finding that only 13% of organizations believe their agent governance is adequate. Ask whether your teams know which agents run on laptops and inside SaaS tenants, which hold long-lived API keys, and who can revoke them. Because Okta only plans Agent Gateway and endpoint discovery for Q3, and the expanded kill switch for Q4, confirm delivery dates before designing controls around them.
CFOs and procurement leaders should plan for agent security to carry a cost. McKinnon's 25% to 37.5% uplift came from a few sizable deals he called a small sample, but if it holds more broadly, agent security will become a material line on identity renewals. Ask vendors whether agent identities are priced per agent, per action or as a platform add-on, and negotiate caps before agent counts grow. Ask, too, which alliance commitments your specific tools will implement; a shared architecture helps only if your products adopt it.
Boards should ask management who owns AI agents as a risk category: security, IT or the business units deploying them. At Oktane, ServiceNow vice president Bhakti Pitre argued that deciding when to shut down an agent requires business context rather than a simple on-off switch, SiliconANGLE reported. That points to the need for clear decision rights before an incident, not during one.
The bigger picture
Identity vendors are racing to become the control point for AI agents, and the lines between partners and rivals are blurring. ServiceNow, which Okta executives named as a competitor, is also working with Okta on securing agent session tokens and identities, according to SiliconANGLE. For buyers, the Blueprint Alliance could reduce integration work across security tools, but beyond the published reference architecture, its signal sharing and interoperability work are still commitments.
What’s next
Watch for the alliance's first joint interoperability results and reference integrations, and for Okta to ship Agent Gateway and Shadow AI Agent Discovery on its stated Q3 timeline. Okta's next quarterly report should show whether agent products are lifting deal sizes as McKinnon suggested.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
Companies in this story
