Anticipated executive order could give NSA a role in voluntary AI model testing
Anticipated executive order could give NSA a role in voluntary AI model testing
Publish Date: 2026-05-20 10:31:00
Source Domain: www.nextgov.com
- White House planning an AI executive order featuring a voluntary information-sharing framework for safety testing of AI models.
- NSA expected to lead classified testing for government, coordinating with developers for pre-deployment evaluations.
- Detailed discussions are ongoing, with the potential for changes in directive specifics and timing.
- Administration balancing opposing views on AI regulation and innovation, leading to varying policy approaches.
- Trump administration tends to prefer intelligence community over Commerce Department for model evaluation, showing leadership preference in AI governance.
- Advanced AI models like Anthropic’s Mythos Preview and OpenAI’s GPT-5.5 are fueling current discussions on potential cybersecurity risks.
- Pre-deployment testing aims to identify and mitigate cyber vulnerabilities ahead of AI model release.
- Administration’s stance on AI-focused information-sharing aligns with last summer’s AI Action Plan, advocating for agency cooperation in AI security threat information sharing.
