IBM’s AI Risk Atlas
AI-generated review:
IBM’s AI Risk Atlas offers a comprehensive overview of the multifaceted risks associated with agentic and generative AI systems. The page categorizes threats into areas like fairness, privacy, robustness, governance, and societal impact, making it easier to navigate complex concerns. It highlights how agentic AI-systems that act autonomously toward goals-introduces amplified or entirely new risks. Examples include trust exploitation, function-calling hallucinations, and impacts on human agency and dignity. The documentation also explores traditional AI risks that persist across models, such as data bias and lack of transparency. Each risk is tagged to indicate whether it’s specific to agentic AI, generative AI, or broadly applicable. Overall, the atlas serves as a valuable resource for developers, policymakers, and researchers aiming to build and govern AI responsibly.
Source: https://www.ibm.com/docs/en/watsonx/saas?topic=ai-risk-atlas