DigiCert launches AI Trust Manager with agent passports and a kill switch

DigiCert said DigiCert AI Trust Manager is generally available inside DigiCert ONE. The product issues portable, cryptographically signed “AI Passports” that tie an agent to an accountable owner, attach policy “visas” for what it may access, and support a kill switch that can revoke authority and quarantine an agent when trust changes.

SAFETY desk — same-day company primary that a major internet-trust vendor put shipping cryptographic identity and a kill switch on autonomous agents, not another vague AI-governance whitepaper.

Why the company says this exists: AI agents can access sensitive data, use software tools, write code, make decisions, and complete transactions with limited human involvement. DigiCert says most enterprise security systems were not built to govern autonomous agents moving across applications, clouds, and organizations at machine speed. File that gap as DigiCert’s. This desk did not audit a customer estate.

What an AI Passport is, as the same wire has it: a portable, cryptographically signed credential that verifies an agent’s identity and connects it to an accountable owner. Cryptographic here means a math-backed proof other systems can check, not a shared password. Policy-based “visas” define which systems, data, and actions the agent may access, and how long that authority lasts. The company contrasts that with traditional identity systems that work like employee badges inside one organization. File the passport / visa picture as DigiCert’s. This desk did not issue a passport.

Cross-organization verify, still company: because the passport and its permissions travel with the agent, other systems and organizations can independently verify them without sharing the same identity provider. A receiving organization can deny the agent access inside its own environment. File that verify-without-the-same-directory picture as DigiCert’s. This desk did not sit between two companies.

Kill switch, company: if an agent attempts a prohibited action, DigiCert’s automated kill switch can block it and, under policies set by the agent’s owner or passport issuer, immediately revoke its authority at the source and quarantine the agent before harm occurs. A kill switch here is a control that cuts the agent’s authority. Quarantine means isolate the agent so it cannot keep acting. File that block / revoke / quarantine sequence as DigiCert’s. This desk did not trip a kill switch.

Enterprise jobs named on the same wire, still company: find and identify agents that were purchased, built internally, or already operating, and connect each one to an accountable owner; set clear limits on systems, data, actions, and how long permission lasts; extend trust across organizations so others can verify an agent without using the same identity provider; and respond when trust changes — update credentials, expire permissions, or revoke an individual agent or groups of agents when policy or risk changes. File that four-item list as DigiCert’s. This desk did not inventory a fleet.

Survey, company — DigiCert’s July 2026 survey of 1,001 IT and cybersecurity leaders, not an independent audit this desk ran: 78% said their organizations had experienced an AI-related security incident or identified an AI vulnerability in the previous year. File the 1,001 / 78% lines as DigiCert’s survey. This desk did not rerun the poll.

Named voices on the release: Amit Sinha, chief executive officer of DigiCert; Grace Trinidad, research director, AI Security and Trust, IDC; and Ajitha Choudary, vice president of Global Security Engineering & IAM at UKG. IAM is identity and access management — who is allowed to use which systems. File the names and titles as theirs, via DigiCert. UKG appears as a quoted endorsement on the company wire — this desk is not inventing a customer-logo roster beyond that quote. Their quotes are color only and stay in Sources.

Architecture, still company: AI Trust Manager is part of DigiCert’s broader AI trust architecture, which the company says applies cryptographic verification across AI agents, models, and content, and complements existing identity and cybersecurity systems with a portable foundation for verifying an agent’s identity and authority wherever it operates. File that complement-not-replace frame as DigiCert’s. This desk did not replace a customer identity stack.

Plain English for the rest of the card: cryptographic = a math-backed proof you can verify, not a shared password. AI Passport = DigiCert’s signed credential tying an agent to an owner. visa here = a policy sticker on that passport for what the agent may touch and for how long. kill switch = a control that instantly cuts an agent’s authority. quarantine = isolate the agent so it cannot keep acting. GA = generally available, out of preview. DigiCert ONE = the company’s existing trust platform. MCP = Model Context Protocol, how some agents talk to tools — named only in the IDC quote, not as the center of this product.

PRIMARY here: DigiCert’s 15 Sep 2026 GlobeNewswire company release — Tier A PRIMARY company source, the original record. The AI Trust Manager product page is product-home context, not a second originating newsroom. The DigiCert ONE GA, the AI Passport / policy-visa picture, the cross-org verify-without-the-same-identity-provider line, the automated kill switch and quarantine, the find / limit / extend / revoke job list, the July 2026 DigiCert survey of 1,001 leaders and the 78% incident-or-vulnerability line, the Sinha / Trinidad / Choudary names, the UKG endorsement on the wire, and the complement-existing-identity architecture are company-attributed. Survey figures stay DigiCert survey figures — not an independent audit. NOT claimed: pricing, a customer-logo roster beyond the quoted UKG endorsement, independently verified incident rates, independent penetration-test results, that this desk tested AI Trust Manager, a stock tip, or investment advice. Distinct from the already-filed esentire-atlas-aidr, wso2-agent-manager-ga, aiuc-40m-series-a, rockwell-anthropic-glasswing, exein-270m-physical-ai, hp-zbook-ultra-g3a, simscale-onshape-engineering-agent, and prophix-one-autonomous-planning-agent.