Trendora
vien

DPACT framework

Assess

Techniques

A security blueprint for AI agents focused on Delegation, Policy, Auditability, Context, and Time.

Why it's here

Placed in Assess: 1 article(s) of evidence from 1 source(s), led by framework updates, with 1 in the last 30 days. Confidence 24%. Low accumulated evidence, so it defaults conservatively pending more signal.

Evidence (1)

  • 4InfoQ·9/21/2026framework_update
    Securing AI Agents with Identity, Authorization, and DPACT

    Sahil Agarwal discusses the security challenges of identity, authorization, and access control for AI agents. He presents the DPACT framework—Delegation, Policy, Auditability, Context, and Time—as a way to build bounded, guardrailed agentic systems instead of relying on simple token-based access.