The National Institute of Standards and Technology (NIST), the agency within the United States Department of Commerce responsible for promoting American innovation and industrial competitiveness, issued new cybersecurity guidelines to embrace emerging artificial intelligence (AI) technologies.
The “Cybersecurity Framework Profile for Artificial Intelligence” (NISTIR 8596) provides guidelines for using the NIST Cybersecurity Framework (CSF 2.0) to accelerate the secure adoption of AI and to help organizations prepare for AI adoption while addressing emerging cybersecurity risks associated with the rapid advance of these technologies.
“Regardless of where organizations are on their AI journey, they need cybersecurity strategies that acknowledge the realities of AI’s advancement,” said Barbara Cuthill, one of the profile’s authors.
Year-long Effort
The publication of NISTIR 8596 is part of a yearlong effort carried out by NIST cybersecurity and AI experts. According to NIST, more than 6,500 individuals in the “community of interest” contributed to the profile’s development. The initial concept paper was released in February 2025, following a workshop in April and a series of community-of-interest meetings hosted by NIST last summer.
The new preliminary draft of the profile has been released with a 45-day public comment period. As NIST noted, the profile has three focus areas:
- Securing AI systems: identifying cybersecurity challenges when integrating AI into organizational ecosystems and infrastructure
- Conducting AI-enabled cyber defense: identifying opportunities to use AI to enhance cybersecurity, and understanding challenges when leveraging AI to support defensive operations.
- Thwarting AI-enabled cyberattacks: building resilience to protect against new AI-enabled threats
“The three focus areas reflect the fact that AI is entering organizations’ awareness in different ways,” Cuthill explained. “But ultimately every organization will have to deal with all three.”
The preliminary draft is intended to solicit public feedback to inform an initial public draft, which will further refine the profile. That will include the mapping of additional relevant resources to the CSF, NIST added.
The initial public draft will be released early next year, but the even after it is finalized, the authors will continue developing the profile as necessary to address further advances in AI.
“The Cyber AI Profile is all about enabling organizations to gain confidence in their AI journey,” Cuthill continued. “We hope it will help them feel equipped to have conversations about how their cybersecurity environment will change with AI and to augment what they are already doing with their cybersecurity programs.”
Some Important Guidance is Missing
There are already concerns that the document may have omitted guidance on complex AI systems, warned Melissa Ruzzi, director of AI at cybersecurity provider AppOmni.
“The draft document doesn’t include guidance for complex systems where AI is used in an orchestration form: for example, one AI leading the work for the next AI, or AI agents that use other AI tools like GenAI or machine learning,” Ruzzi explained in an email to ClearanceJobs. “In these cases, the hyperparameters are defined through AI itself and cannot deterministically be controlled as some of the NIST guidance suggests.”
The NIST profile could be seen as a primer on AI as it relates to cybersecurity, more than the final word.
“This is a great guidance for those who don’t have much expertise in AI security. The caveat is that a better explanation about when something is applicable or not is needed,” Ruzzi suggested.
She warned that most of the document’s points are framed as possibilities, underscoring the many ways to implement AI.
“There’s no fully right or wrong way, or one size fits all,” said Ruzzi. “Depending on AI’s business purpose, public interaction and the types of users are key decision factors as to whether each point in the document is applicable or not, in each case.”
AI has advanced considerably in a very short time, and it will continue to do so.
“The guidance is very helpful for those wanting to learn more about AI security considerations and needing a clear list of where to start,” Ruzzi continued, “but it shouldn’t be blindly followed without analyzing applicability to each use case.”
