Make it safe to say yes to AI agents.
AI agents are starting to do real jobs. They book travel, write code, make trades and resolve support tickets. Most security thinking treats that as a threat to contain. We treat it as progress worth protecting. MITRITY exists so organizations can hand real work to agents with confidence, because every action is checked against what the agent is for before it runs.
Each new layer of computing got its own security. Agents are next.
Networks got firewalls, applications got WAFs, and identities got IAM.
Gartner projects the average Fortune 500 enterprise will go from fewer than 15 AI agents in 2025 to over 150,000 by 2028.
Source: Gartner, “Six Steps to Manage AI Agent Sprawl,” April 2026.
Now software acts on its own. It calls APIs, moves data and spends money, and the controls built for people and static code don't fit. Watching agents isn't enough. Someone has to govern them while they act, at machine speed.
That's the layer we're building. It knows what each agent is for, decides in under a millisecond, and is built to say yes far more often than it says no.
Thirty years in security, and a pattern he kept seeing
Joakim has spent three decades building security just as the industry changed shape. He worked on network intrusion detection at Sourcefire (later part of Cisco) and on application security at F5 Networks. Each time, he watched a new layer of computing arrive before its safeguards did.
Before MITRITY he founded Baffin Bay Networks, a threat-protection platform based in Stockholm, and led it until Mastercard acquired it in 2023.
MITRITY comes from spotting the same pattern one platform shift earlier. Autonomous agents are the new layer, and they can't ship safely without runtime governance. This time the safeguards arrive with the shift, not years after it.
Interested in what we're building?
We'd love to hear from you, whether you're putting agents to work, researching the field or thinking about joining us.