Agentic artificial intelligence, defined as AI systems that can operate autonomously with minimal human oversight, is rapidly evolving beyond current prompt-based models. According to experts at the Special Competitive Studies Project (SCSP), a nonprofit and nonpartisan initiative, these systems can independently set goals, create plans, and execute multi-step tasks, marking a significant shift in AI capability. Ylli Bajraktari, president of SCSP, emphasized in a recent newsletter that a self-accelerating loop where AI helps build better AI could cause capability development to far outrun projections.
The implications for business and global security are profound. Bajraktari warned that adversaries could deploy agentic AI to navigate complex bureaucratic systems, identify exploitable vulnerabilities, and act without clear attribution, representing a qualitative expansion of adversarial capability in areas where governance is weakest. This could enable coercion, espionage, and influence operations, making it critical for leaders to understand and govern these systems effectively.
However, SCSP experts argue that effective governance hinges not on the AI model itself but on the scaffolding built around it. This scaffolding includes connectors to bridge models to real-world infrastructure like email and financial platforms, memory for learning across interactions, planning capabilities to break objectives into tasks, permission structures defining access, and guardrails that set refusal limits such as spending caps or human sign-offs. Accountability remains a challenge, with current frameworks falling short in three key ways: responsibility is untraceable when AI acts on a user's behalf, evaluations focus only on task completion rather than safety or harm, and agentic AI builds personal profiles that may include sensitive data on behavior patterns and preferences.
Despite these challenges, SCSP emphasizes that agentic AI should not be feared or deferred. Institutions that prioritize understanding, shaping, and governing this technology will determine their competitive position and influence the global environment in which agentic AI operates. For leaders in business and technology, the message is clear: proactive engagement with governance structures is essential to harness agentic AI's potential while mitigating risks. Visit scsp.ai to explore how the United States can pursue effective governance of agentic AI.
