Millions of AI Agents to Test Enterprise Controls
The proliferation of autonomous AI agents makes it necessary for companies to transition from data access security to complex runtime authorization and accountability management.
Entrust Chief Operating Officer Anudeep Parhar stated that autonomous AI agents will push corporate control mechanisms to their absolute limits, and the focus will shift from authentication to authorization.
The Rise of AI Agents
In recent years, companies have focused on the secure use of artificial intelligence for employees to retrieve information, summarize documents, and generate recommendations. This new era, known as Agentic AI, fundamentally changes this model as software gains the ability not just to answer questions, but to write to systems, make decisions, and execute transactions.
Authorization and Accountability
Speaking at a PYMNTS event session, Anudeep Parhar stated that organizations could have thousands or even millions of agents acting on behalf of customers and employees. This situation requires companies to delegate authority, a step far more sensitive than simple data access.
Machine-Speed Operational Architecture
For agentic commerce to scale successfully, the infrastructure needs to answer certain questions at machine speed. Issues such as whether a transaction is authorized, whether the target has been reached, and who is responsible for the authorization are of vital importance.
From Human Speed to Machine Scale
Traditional controls were built for human speed, but because agents can operate continuously and across systems, it becomes impossible for every decision to be reviewed by a human. This necessitates the enforcement of existing policies at agent scale.
Technical Capacity and Corporate Authority
A security agent may patch a critical system, but technical capacity alone does not equate to corporate authority. Organizations maintain certain approval thresholds and human oversight mechanisms for financial transfers or infrastructure changes based on their risk profiles.
Auditors and Regulatory Frameworks
It is emphasized that when an erroneous transaction with major consequences occurs, the answer to be given to auditors cannot simply be to blame the agent. Artificial intelligence is evolving from a pure technology issue into a fundamental identity and authorization problem.