On May 21, 2026, the New York State Department of Financial Services (“NYDFS”) issued two related industry letters: an advisory to CISOs of NYDFS-regulated entities addressing heightened cybersecurity risks associated with certain frontier AI models (the “Advisory”), and related guidance on measures that NYDFS-regulated entities should consider in a heightened cyber threat environment, including but not limited to the use of AI (the “Guidance”). While the Advisory and Guidance are directed at NYDFS-regulated entities, they lay out what may become the standard of care and reasonable practices relevant to any organization managing cybersecurity risks in connection with heightened threat environments. The risk posed by adversaries misusing Frontier AI Models and recent attacks by groups like ShinyHunters, Silent Ransom Group, and TeamPCP, underscore the need for organizations to reassess their readiness for today’s threat environment.
The Advisory and Guidance explicitly do not establish new legal requirements for NYDFS-regulated entities. But they do offer practical recommendations that entities may consider in connection with their existing cybersecurity controls and risk-management practices required by NYDFS’s cybersecurity regulation (“Part 500”).
In this blog post, we summarize the Advisory and Guidance and offer practical takeaways for NYDFS-regulated entities, and other organizations more broadly, seeking to navigate the latest cybersecurity risks associated with AI.
The Advisory and Guidance Explained
The Advisory addresses “Frontier AI Models,” which NYDFS describes as models that amplify the potency, scale, and speed of identifying vulnerabilities and exploits in information systems. We previously wrote about a specific Frontier AI Model, Anthropic’s Mythos, in discussing practical considerations for responding to a heightened cybersecurity landscape.
The Advisory and Guidance provide a high-level, non-exhaustive list of recommended measures for NYDFS-regulated entities to consider incorporating into their existing cybersecurity program. The recommendations fall into three categories: (1) reducing attack surface, (2) improving threat detection and readiness, and (3) improving resilience and response.
1. Reducing Attack Surface
An entity’s “attack surface” broadly refers to the various points an unauthorized actor could exploit to gain access to its systems or data. Managing that attack surface has become an increasing industry focus as attack vectors grow more sophisticated, including as a result of Frontier AI Models. Given the speed with which AI tools may enable threat actors to identify and exploit system vulnerabilities, NYDFS recommends quick identification and remediation of known exploited vulnerabilities in firmware, hardware, and software.
To reduce attack surface, the Guidance recommends both AI-specific controls and broader cybersecurity hygiene that entities should consider, some of which are already either required by Part 500 or highly encouraged by prior NYDFS guidance to Part 500. For example, AI-specific measures include secure coding practices for AI-assisted development, while broader cybersecurity recommendations include privileged access review, network access limitations (e.g., network segmentation), and phishing-resistant multi-factor authentication.
2. Improving Threat Detection and Readiness
The Guidance emphasizes being able to quickly identify vulnerabilities to improve organizational threat detection and readiness in an increasingly complex cyber threat environment. A key recommendation highlighted in the Advisory is coordinating with third-party service providers to identify third-party dependencies, address significant vulnerabilities, and manage related action plans. Other recommendations include establishing logging and alerting upon anomalous or suspicious activity, qualified human review of threat intelligence and relevant guidance, and employee education on preventing, detecting, and responding to the latest cyber threat campaigns. While many of these recommendations are already part of most organizations’ cybersecurity programs, the Department is emphasizing the importance of staying up to date with the latest threat trends.
3. Improving Resilience and Response
Lastly, the Guidance offers recommended measures to improve an organization’s ability to respond to cyber threats. One key recommendation is to review and test operational resilience procedures for critical functions, information systems, and nonpublic information, which is consistent with existing obligations under Part 500. Further, NYDFS highlights the continued significance of establishing communication strategies in the event of system and service disruptions, and of monitoring for suspicious financial transactions.
Practical Takeaways
1. Maintain and Refresh Employee Training and Communications. As employees remain a common target for threat actors, and technological capabilities leap forward through the availability of Frontier AI Models, consider a risk-calibrated cadence for distribution of relevant guidance and employee training to help improve organizational readiness for responding to AI-enabled attack vectors.
2. Map and Manage Third-Party Dependencies. Mapping existing vendor and third-party dependencies is key to understanding your organization’s attack surface. Effective third-party risk management may also include incorporating appropriate contractual terms for new vendor contracts during vendor diligence to maintain adequate cybersecurity controls for your systems and confidential data.
3. Reassess Patch Prioritization and Escalation. As we wrote when Mythos was first unveiled, consider revisiting your patching protocol and framework. In line with the risk-based framework of Part 500, this requires balancing the reality of business disruption during the patching process against the speed with which threat actors can uncover and exploit vulnerabilities using AI tools.
4. Increase Segmentation, Strengthen Monitoring, and Minimize Data Retention. NYDFS’s Advisory and Guidance underscore the importance of network segmentation and monitoring in the current threat environment. Recent enforcement activity also reflects NYDFS’s continued focus on data minimization: limiting the amount of sensitive data that organizations retain, and how long they keep it. Organizations should consider incorporating these best practices into their cybersecurity programs.
5. Establish Comprehensive Asset Visibility to Address Shadow IT. Many of the measures NYDFS recommends for reducing attack surface depend on an organization’s ability to maintain a current inventory of the assets operating in its environment. That inventory should include not only traditional hardware, software, and cloud assets, but also AI tools and agents deployed or used within the organization’s systems. Improved visibility can help security teams address shadow IT and shadow AI risks before they expand the organization’s attack surface.
6. Use the Advisory as a Board and Senior-Management Conversation Starter. As companies respond and adapt to the new AI-charged vulnerability landscape, expectations around “reasonable security” and similar concepts will evolve. Accordingly, Legal and Compliance teams should consider staying connected with their CISOs and information-sharing councils to understand how other organizations are adapting and what reasonable security looks like. Internally, the latest NYDFS Advisory and Guidance give organizations a concrete basis to revisit whether their budgets, remediation timelines, legacy-system plans, vendor oversight, and incident response procedures remain adequate for an AI-amplified threat environment.
***
To subscribe to the Data Blog, please click here.
The cover art used in this blog post was generated by ChatGPT.
The Debevoise STAAR (Suite of Tools for Assessing AI Risk) is a monthly subscription service that provides Debevoise clients with an online suite of tools to help them responsibly fast-track their AI adoption. Please contact us at STAARinfo@debevoise.com
