Your AI Agents Are Guessing at Scale: Permissions Decide the Damage
July 29, 2026 · BleepingComputer · Severity: MEDIUM
This sponsored article by Token Security argues that traditional security approaches — prompt filtering, behavioral guardrails, and static IAM policies — are fundamentally inadequate for securing AI agents because agents operate probabilistically, improvising their next actions rather than following predictable workflows. The author contends that the only effective control plane for agentic AI is identity, since every action an agent takes runs through an identity (service accounts, API keys, OAuth grants, cloud roles, or tokens). Traditional IAM asks what an identity can access, but for agents, intent is the missing dimension — what the agent should be able to access to accomplish its specific purpose. With non-human identities already outnumbering human identities 109-to-1 according to Palo Alto Networks, and agents that never sleep, don't use MFA, and are rarely retired, the article argues organizations need to discover every agent and its identities, understand what each can access, and enforce intent-based permissions automatically across the full lifecycle.
Key Takeaways
- Traditional security fails for probabilistic agents — AI agents improvise actions unpredictably, making prompt filtering and behavioral guardrails too late to prevent damage.
- Identity is the only cross-system control plane — Every agent action runs through an identity (service account, API key, OAuth grant), making identity the consistent layer for enforcement.
- Intent-based IAM adds the missing dimension — Beyond what an identity can access, intent defines what it should access for its specific purpose, enabling operational least privilege.