Every agent – whether an in-house agent, a third party or otherwise – receives a cryptographic token that is verified at runtime; if it strays outside its authorised scope, it may be redirected, restricted or cut off completely.
On October 15, ServiceNow presented AI Passport, an identity mechanism for AI agents in companies. The principle: each agent, whether developed on the platform or provided by a third party, is assigned a cryptographically signed credential that describes the data, tools and actions to which he or she is entitled, as well as the person who is responsible for them. This credential is checked during execution, and not just at the time of deployment.
As long as the agent remains within its perimeter, it acts freely. If it tries to get out of it, graduated responses apply: redirection, restriction of rights, recontextualization, or even immediate shutdown via a "kill switch". All these interventions are logged for auditing. The system is based on open specifications, so that other identity providers or governance tools can issue and verify these credentials, including from one step to the next in multi-agent chains.
Nvidia launch partner
Nvidia is a design partner: AI Passport is based on its Open Agent Safety platform, which includes the OpenShell secure execution environment and the Sentry framework. The package integrates with ServiceNow's governance console, renamed the Super Intelligence Control Tower, which adds shadow AI detection, more than 400 connectors and cost tracking per user and per department. The integrator LTM says it already manages 1,500 agents, each reporting to a manager.
Read the full article on the original source
Go to the source →