The Year(s) of the AI Agents: Opportunities and Challenges in Depth

2025 has been called the “year of the AI agent,” as AI agents are currently poised to revolutionize how people interact with AI and the future of work and productivity across industries. Their ability to autonomously gather and analyze data, make decisions, and take actions in the real world positions them as critical tools in the modern technological landscape. However, the rapid technological advancements involved in developing and deploying AI agents are likely to test current AI technical and governance frameworks, including due to the importance of establishing effective guardrails and monitoring systems for AI agent workflows to mitigate operational and legal risks. The future of AI safety and governance in an agentic world must focus on what types of testing, tools, frameworks, and methodologies we’ll need to actually understand and govern agentic systems and architectures. We need to start preparing for the next chapter of AI safety and governance with AI agents, including how to understand, measure, and mitigate risks involving AI agents’ interaction with digital and physical worlds and multi-agent interactions.