The Battle to Govern Agentic AI

(NewsUSA) - Agentic artificial intelligence (AI) refers to AI systems that can operate autonomously with minimal human oversight. Unlike existing AI systems that generate responses based on prompts, evolving agentic AI can independently set goals, create plans, and execute multi-step tasks.

“AI is beginning to help build better AI,” according to experts at the Special Competitive Studies Project (SCSP), a nonprofit and nonpartisan initiative with a goal of making recommendations to strengthen America's long-term competitiveness in artificial intelligence.

If a self-accelerating loop in which AI capability improves and AI development compounds, the pace of capability development will far outrun projections, said Ylli Bajraktari, president of the SCSP, in a recent newsletter about agentic AI.

“An agent that can navigate complex bureaucratic systems, identify exploitable vulnerabilities, and act without leaving a clear attribution trail represents a qualitative expansion of adversarial capability,” Bajraktari emphasized.

In terms of global security, the United States should be aware that our adversaries will deploy agentic AI systems in areas where governance is weakest. Agentic AI systems could be used for coercion, espionage, and influence.

Contrary to what many policymakers might think, effective governance of agentic AI involves not the AI model that is the engine, it is the scaffolding built around it, the SCSP experts explained. This scaffolding includes:

Connectors to bridge the model to real-world infrastructure, such as email, booking systems, and financial platforms.

Memory that allows the AI system to learn and adapt across interactions over time.

Planning capabilities that break large objectives into smaller tasks, identify failure, and navigate obstacles without human intervention.

Permission structures that define what the system can access and act upon.

Guardrails that determine what the system will refuse to do (such as spending limits or human sign-offs).

Accountability remains a challenge, and governance is falling short in three key ways, according to SCSP. First, responsibility is untraceable when using AI; there is not way to determine who authorized what if an AI agent is acting on your behalf. In addition, current frameworks don’t ask whether an AI agent performed a task safely or whether it caused harm, only that the task was completed. Finally, agentic AI builds personal profiles that may include data more sensitive than an individual wants by accumulating information on patterns of behavior, preferences, and inferences.

Despite these challenges, agentic AI is not a technology to be feared or deferred, the SCSP experts emphasize. Institutions that prioritize understanding, shaping, and governing agentic AI will determine their own competitive position, and also impact the character of the environment in which agentic AI operates globally.

Visit scsp.ai to learn more about how the United States should pursue effective governance of agentic AI.