Okta

Your AI agents already have access. Can you prove what they can do? The EU AI Act is about to make “we think so” a very expensive answer. 81% of CISOs are concerned about AI access they can’t control. The fix isn’t more compliance theater. It’s identity: knowing which agents exist, what they can access, and who’s accountable when they act. The latest Access Granted breaks down what the EU AI Act means for enterprise security—and how to build the governance foundation before 2027.

AI agents don’t just spend tokens when they use a tool. They spend tokens deciding what tools they could use. That means every extra tool exposed to the model adds cost before any real work happens. We call that the tool tax. By scoping tools to what an identity is actually allowed to use, Okta can reduce the number of visible tools — and the tool-selection cost tied to them — by over 90% in some scenarios. → Fewer tools in the prompt → Lower tool-selection cost → Smaller blast radius if an identity is compromised Learn more: https://lnkd.in/eqh_qENv

Think your AI agent rollout is moving at a manageable pace? Think again. 📈 Speaking on Bloomberg, our CEO Todd McKinnon shared a concrete story from a recent Okta for AI Agents deal. When the evaluation started, Okta's tools detected 50 agents in the customer's environment. Just a few weeks later, that number had exploded to 1,500 agents. We're officially out of the chatbot phase and into the era of autonomous agents. These agents are incredibly powerful and proliferating fast, but without proper identity management, they introduce massive, unforeseen risks.

Our Q2 FY27 financial results are in. Identity is the foundation to securing the agentic future, and we’re helping our customers build for it today. Here are just a few of the things we're proud of from this past quarter: → Total revenue +11% YoY → Total cRPO +14% YoY → Free cash flow $226M (28% margin) → Acquired Permiso Security → Expanding XAA with 25+ integrations Tune in at 5 pm ET today for more: https://lnkd.in/esiqyvkm

AI agents can move fast. That’s the promise. But if you can’t see them, govern them, or stop them, that speed becomes risk. Harish Peri is getting ready to talk AI security at Oktane 2026. September 22–24 | Las Vegas | Register now https://lnkd.in/eEr_ZiY4

AI agents are being built everywhere. Gemini. Microsoft. Workday. DataRobot. Glean. LangSmith. That’s not the problem. The problem is when every platform becomes its own agent island. Separate registry. Separate policy layer. Separate visibility gap. Security can’t govern agents it can’t inventory → https://lnkd.in/gfkxHW_a

Answering unknown calls? Risky. Letting unknown AI agents run? Even riskier. 📞😳 Know exactly who’s on the line. Okta secures AI: https://bit.ly/4vTsMUV McLaren Racing #McLarenF1

AI teams don’t wait for governance to catch up. They build. That’s great for velocity. It’s risky when every agent platform creates its own registry, policy layer, and visibility gap. Okta’s expanded AI Agent Import library helps bring agents from supported builder platforms into one identity control plane—so security teams can govern ownership, access reviews, lifecycle controls, and deactivation centrally. Say yes to the agent platform, not the governance blind spot: https://lnkd.in/gfkxHW_a

Explaining the secure agentic future is hard. Explaining it without saying “agentic” 37 times? That’s why Shannon Sullivan Duffy gets the CMO title. See what she’s preparing for Oktane 2026. September 22–24 | Las Vegas Register now: https://lnkd.in/eEr_ZiY4