US Federal News Bureau
Written by: Tathagata Sen
Updated 11:29 PM EST, November 20, 2026

The U.S. Government Accountability Office (GAO) has outlined 16 policy options for addressing the cybersecurity risks and opportunities created by artificial intelligence, including stronger AI testing, agent identity management and human authorization for consequential AI-driven security actions, according to a report released November 20.
The report, AI and Cybersecurity: Technologies Can Magnify Cyberattacks and Enhance Cyber Defense, examines how AI is changing both offensive cyber operations and defensive capabilities.
GAO found that generative AI can improve the precision, speed and scale of attacks through convincing phishing content, deepfakes and automated attack techniques. Agentic AI is also emerging as a tool for carrying out multiple stages of an attack with minimal human input.
One example cited in the report involved a June 2026 incident in which a threat actor used multiple AI agents to build a malware-testing framework. The agents identified attack techniques from security research, mapped them to cybersecurity frameworks, tested malicious code against different security products and reported the results.
AI is also helping defenders analyze network traffic, identify vulnerabilities in source code and process large volumes of security information. However, GAO found that there are no widely accepted criteria for evaluating the offensive and defensive capabilities of generative AI systems.
Experts consulted by GAO identified data privacy and quality, outdated AI governance frameworks, technical limitations and difficulties integrating AI tools among the challenges facing cyber defenders. Fragmented regulatory approaches and insufficient data sharing were also identified as obstacles.
To address these issues, GAO identified four potential policy goals: closing gaps in cybersecurity guidance and practices, addressing AI-related research gaps, governing AI tools throughout their lifecycle and strengthening the cybersecurity workforce.
The 16 policy options include expanding AI testing environments, improving research into agent identity management, validating AI tool outputs before and during deployment, and requiring human authorization for consequential AI-initiated cybersecurity actions. Other options include establishing fallback capabilities for critical functions and taking a risk-based approach to acquiring AI products.
GAO presents these as alternatives for policymakers, including Congress, government agencies, standards organizations and industry, to consider. They are not mandatory requirements established by the report.
For chief data officers (CDOs), the report highlights how data quality, privacy and information sharing affect the reliability and security of AI-enabled operations. Cybersecurity tools depend on data from across networks and systems to identify threats and support decisions.
Poor-quality or fragmented information can undermine those capabilities, while insufficient data sharing can limit the visibility available to defenders.
The proposed focus on governing AI tools throughout their lifecycle also has implications for data and AI governance teams. Organizations need to assess the data used by security tools, validate their outputs and establish accountability for consequential actions taken with AI assistance.
Agent identity management is another emerging concern as autonomous systems gain the ability to interact with networks and execute complex tasks.
These priorities align with the broader need to operationalize AI governance, particularly where AI systems operate within security-critical environments. CDOs, working with cybersecurity and technology leaders, can help ensure that data access, quality, privacy and accountability controls keep pace with the adoption of AI-powered tools.