NIST CSF 2.0

At minimum: the list of approved AI tools and the firm accounts staff must use, what data may and may not go into each tool, who has access, how use is logged and reviewed, and how staff request a new tool or an exception. Each rule is stronger when it names the technical control that enforces it, such as web filtering or DLP.

Microsoft states that Privileged Identity Management requires Microsoft Entra ID P2 or Microsoft Entra ID Governance. Business Premium includes Microsoft Entra ID P1, so a firm on Business Premium would need an add-on license to use just-in-time admin role activation. Without it, a firm can still keep standing admin assignments few and remove roles manually when they are no longer needed.

An emergency access account, often called a break-glass account, is a cloud-only account that holds the Global Administrator role permanently and is not assigned to any one person. It exists so the firm can still administer the tenant if normal admin sign-in fails, for example during an MFA outage. Microsoft recommends keeping two of them and reserving them for emergencies only.

Microsoft recommends assigning the Global Administrator role to fewer than five people, and the admin center shows an alert at five or more. Microsoft also recommends two cloud-only emergency access accounts permanently assigned the role, kept for when normal sign-in fails. Everyone else who does admin work can use a narrower built-in role such as User Administrator or Exchange Administrator.

No. NIST describes the AI RMF as voluntary, and no SEC, FINRA or FTC rule requires a private firm to adopt it. Firms use it as a shared vocabulary to show examiners, auditors and clients how their AI policy addresses recognized risks. NIST has said AI RMF 1.0 is being revised as part of the White House AI Action Plan.

Version 2.0, released in February 2024, added a sixth function, Govern, covering strategy, roles and oversight. It also expanded the framework's intended audience from critical infrastructure to organizations of every size and gave more weight to supply chain risk.

NIST CSF 2.0 has six functions: Govern, Identify, Protect, Detect, Respond and Recover. Govern was added in version 2.0, published February 26, 2024, and sits at the center because it informs how a firm carries out the other five. Identify, Protect, Detect, Respond and Recover describe what the security program does day to day.

The biggest change is the new Govern function. CSF 1.1 had five functions: Identify, Protect, Detect, Respond and Recover. CSF 2.0, published February 26, 2024, keeps those and adds Govern at the center, covering risk strategy, leadership accountability, policy, oversight and supply chain risk management. It is written for organizations of every size and sector.

A Community Profile is a baseline of CSF outcomes published for organizations with shared interests, such as a sector or a threat type. Examples on NIST's site include the Ransomware Risk Management Profile and the Cyber Risk Institute's Financial Sector Profile. A firm can copy one into its own Target Profile and adjust the priorities.

Partly. CSF 2.0 covers the security outcomes around an AI tool, such as inventories, access control, data protection, monitoring and incident response. AI-specific risks like output validity and third-party models sit in the AI RMF. NIST's Cyber AI Profile, released as a preliminary draft in December 2025, is meant to connect the two through CSF 2.0 functions.

Under NIST CSF 2.0, a Current Profile records the cybersecurity outcomes an organization is achieving today and how fully. A Target Profile records the outcomes it has selected and prioritized as goals, including expected changes such as new requirements or technology. Comparing the two shows the gaps, which become a prioritized action plan.

No. CSF 2.0 is a voluntary set of outcomes, and compliance is measured against each rule's own text. A crosswalk shows where one control or document can serve as evidence for several rules. It does not cover requirements that exist only in a rule, such as the 30-day customer notice in Reg S-P or annual board reporting under the FTC Safeguards Rule.

Govern has six categories: Organizational Context (GV.OC), Risk Management Strategy (GV.RM), Roles, Responsibilities, and Authorities (GV.RR), Policy (GV.PO), Oversight (GV.OV), and Cybersecurity Supply Chain Risk Management (GV.SC). Together they cover how leadership sets risk tolerance, assigns accountability, approves policy, reviews results and manages vendor risk.

Generally, yes. GV.SC covers suppliers and other third parties that provide technology products and services, and an IT provider with administrative access to a firm's systems usually ranks near the top when GV.SC-04 asks that suppliers be known and prioritized by criticality. That makes the provider's contract terms, breach notice commitments and evidence of its own security practices central to a GV.SC program.

GV.SC is the Cybersecurity Supply Chain Risk Management category in the Govern function of NIST CSF 2.0, published February 26, 2024. Its ten subcategories, GV.SC-01 to GV.SC-10, cover knowing and ranking suppliers by criticality, setting security requirements in contracts, due diligence before onboarding, monitoring suppliers over the relationship, including them in incident plans, and planning for the end of the relationship.

NIST does not set a number. SP 800-55 Volume 1 advises that with quantitative metrics, at least at the start, less is often more. For a small firm, two or three metrics per CSF 2.0 function, around a dozen in total, is a practical starting point, as long as each has an owner, a data source and a target.

No. The NIST Cybersecurity Framework is voluntary guidance, and no regulator requires a private professional firm to adopt it by name. Firms use it as a common structure for their security program because its six functions map well to rules such as the FTC Safeguards Rule, Regulation S-P and state reasonable security laws.

Yes. NIST publishes a free CSF 2.0 Organizational Profile template as a spreadsheet that places Current and Target Profiles side by side so gaps are easy to see. NIST Special Publication 1301, the quick-start guide for Organizational Profiles, explains how to scope, complete and use it.

No. CSF 2.0 describes outcomes and leaves the choice of measures to each organization. It does expect performance to be reviewed: subcategory GV.OV-03 calls for cybersecurity risk management performance to be evaluated and reviewed for adjustments, and Section 5.1 describes giving executives key performance and key risk indicators. NIST SP 800-55 offers guidance on choosing measures.

NIST CSF 2.0 organizes cybersecurity outcomes into six Functions: Govern, Identify, Protect, Detect, Respond and Recover. Govern was added in version 2.0 and covers how leadership sets risk strategy, roles, policy and oversight. Earlier versions of the framework had five Functions, which is why some sources still describe five pillars.

NIST CSF 2.0 defines four Tiers: Partial (Tier 1), Risk Informed (Tier 2), Repeatable (Tier 3) and Adaptive (Tier 4). They describe how rigorous a firm's cybersecurity risk governance and management practices are, moving from ad hoc, case-by-case decisions toward formal, firm-wide policy and, at Tier 4, risk management that adapts continuously and is monitored alongside financial risk.

No. NIST says Tiers should complement a firm's risk management method, and that moving to a higher Tier is encouraged when risks or mandates are greater or when a cost-benefit analysis shows a feasible, cost-effective risk reduction. A firm can set different target Tiers for different Functions, based on its risks, obligations and budget, and revisit them each year.

Leadership. NIST SP 1302, the quick-start guide for using the CSF Tiers, says Tier selection is generally performed by organization leadership. It suggests choosing Tiers at the Function or Category level and weighing current practices, threats, legal and regulatory requirements, business objectives, supply chain needs and resource constraints. The IT provider supplies evidence; the partners or board make the call.

No. NIST released the Cyber AI Profile (NIST IR 8596) as a preliminary draft on December 16, 2025, and public comments closed January 30, 2026. NIST said the comments would inform an initial public draft planned for 2026. As of late October 2026, NIST's publication page still lists only the preliminary draft, so its priorities and wording may change.

NIST has not announced a date. After comments on the December 2025 preliminary draft closed on January 30, 2026, NIST said it planned to release an initial public draft in 2026, which comes before a final version. As of October 24, 2026, that initial public draft has not been posted, so plan to revisit your AI controls when it appears.

No. The profile is voluntary, and the draft says its proposed priorities may be higher or lower for individual organizations. A small firm can start with the Govern and Secure items that match the AI tools it actually uses, such as naming an owner for each tool and reviewing AI suppliers, and add more as its AI use grows.

The AI Risk Management Framework addresses the responsible use of AI broadly, including fairness and transparency. The Cyber AI Profile is narrower: it applies the outcomes of NIST CSF 2.0 to the cybersecurity risks of AI systems, AI-enabled attacks and AI used for defense. NIST says the profile is not meant to replace existing frameworks, so a firm can use both.

Yes. NIST SP 800-66 Revision 2, published in February 2024, maps the HIPAA Security Rule's standards and implementation specifications to NIST Cybersecurity Framework Subcategories and to SP 800-53 Rev. 5 controls. NIST hosts the mappings in its Cybersecurity and Privacy Reference Tool, where they can be browsed and searched.

NIST SP 800-55 Volume 1 defines measures as quantifiable and objective values that result from measurement. Metrics are measures and assessment results designed to track progress, facilitate decision-making and improve performance. In short, a measure is a number, and a metric is a number or result you use to steer the program.

A common example at a financial firm is a portfolio accounting server that cannot run current endpoint protection. The firm records the waived control, isolates the server on its own network segment, adds monitoring, has the CCO or Qualified Individual approve the arrangement, and sets an expiry date tied to the vendor's upgrade.

A usable record names the policy section being waived, the specific system or person covered, the business reason, the risk and how it is accepted, the compensating controls, the approver's name, title and date, and a fixed expiry date. NIST CSF 2.0 subcategory ID.RA-07 describes the same goal: exceptions that are managed, assessed for risk impact, recorded and tracked.

CIS Safeguard 5.4 calls for restricting administrator privileges to dedicated administrator accounts and doing email, browsing and everyday productivity work from a non-privileged account. In Microsoft 365, that usually means each admin keeps a normal licensed account for daily work and a second cloud-only account, protected by MFA, that holds the admin role. A phishing email then reaches an account that cannot change tenant settings.

Most AI acceptable use policies prohibit entering client, customer or other confidential data into unapproved tools, signing up for AI services with personal accounts, installing unapproved AI browser extensions, and sending AI-generated content to clients without the review the firm requires. Each prohibited use should map to a control, such as web filtering, DLP or extension restrictions, that blocks or flags it.

Automation bias is the tendency to accept output from an automated system without enough scrutiny. NIST's Generative AI Profile (NIST AI 600-1) describes it as excessive deference to automated systems and lists over-reliance among the risks of generative AI. In an accounting firm it shows up when staff accept an AI calculation or research answer because it looks polished, without tying it to source documents.