Part 500’s September 2018 controls decided how much a breach could expose

New York’s August 2018 reminder flagged five cybersecurity controls due by September 4: disposal, encryption, user monitoring, audit trails and application security. Together they set what an intruder would find, read and leave behind, and how well a firm could reconstruct it.

Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.

Two-minute summaryThe dek and the first sentence of each section, quoted from the article

New York’s August 2018 reminder flagged five cybersecurity controls due by September 4: disposal, encryption, user monitoring, audit trails and application security. Together they set what an intruder would find, read and leave behind, and how well a firm could reconstruct it.

  1. The week On Wednesday, August 8, 2018, the Superintendent of the New York State Department of Financial Services reminded the entities covered by the state’s cybersecurity regulation, 23 NYCRR Part 500, that the regulation’s third transitional period would end on September 4, 2018.
  2. What is kept: disposal as a limit on exposure The first part concerns data that should no longer exist.
  3. What is readable: encryption and the CISO’s exception The second part concerns the form of the data that remains.
  4. What is visible: monitoring and the audit trail The third part concerns the record of access.
  5. What is built: application security as a design control The fourth part concerns the software through which data is accessed.
  6. Objections, and where the argument is weakest The argument here is that the eighteen-month provisions were loss-limiting controls that set the size of a breach.
  7. With hindsight The regulation was amended.

The week

On Wednesday, August 8, 2018, the Superintendent of the New York State Department of Financial Services reminded the entities covered by the state’s cybersecurity regulation, 23 NYCRR Part 500, that the regulation’s third transitional period would end on September 4, 2018. From that date, the Department said, banks, insurance companies and other financial services institutions it regulated would have to comply with several additional provisions of the regulation.

The release listed them. Covered entities would need an audit trail designed to reconstruct material financial transactions sufficient to support normal operations in the event of a breach. They would need policies and procedures to ensure secure development practices for staff who develop applications for the entity. They would have to implement encryption to protect nonpublic information held or transmitted by the company. They would need policies and procedures for the secure disposal of information no longer necessary for business operations. And they would need a monitoring system that included risk-based monitoring of everyone who accesses or uses the company’s information systems or nonpublic information. The release also listed annual reporting to the board by the chief information security officer, although the regulation itself had put that report, section 500.04(b), in an earlier, one-year group. The Superintendent’s statement described the new protections as including encryption, access controls and audit trails.

The release ended with a reminder about the next deadline. Entities that used third-party service providers had to evaluate the risk those providers posed to the security of their systems and data, and ensure those systems and data were protected, by March 1, 2019.

The dates came from the regulation itself. Part 500 took effect on March 1, 2017. Its transitional periods section gave covered entities 180 days to comply generally, one year for a first group of provisions, eighteen months for sections 500.06, 500.08, 500.13, 500.14(a) and 500.15, and two years for section 500.11 on third-party service providers. Apart from the board report, the controls the release listed were that eighteen-month group.

Those five sections had specific terms. Section 500.06 required systems designed to reconstruct material financial transactions and audit trails designed to detect and respond to cybersecurity events, with records kept for at least five years and three years respectively. Section 500.08 required written procedures for secure development of in-house applications and for evaluating the security of externally developed ones, reviewed by the CISO or a qualified designee. Section 500.13 required policies for the periodic secure disposal of certain nonpublic information no longer necessary for business purposes, except where law required retention or targeted disposal was not reasonably feasible. Section 500.14(a) required risk-based controls to monitor the activity of authorized users and detect unauthorized access to, use of or tampering with nonpublic information.

Section 500.15 required controls, including encryption, to protect nonpublic information in transit over external networks and at rest. Where a covered entity determined that encryption was infeasible, it could instead use effective alternative compensating controls reviewed and approved by its CISO, and the CISO had to review the feasibility of encryption and the effectiveness of those controls at least annually. The regulation’s limited exemption for small covered entities exempted them from several sections, including 500.06, 500.08, 500.14 and 500.15, but not from 500.13.

The September 2018 deadline is easy to read as a list of five controls with one due date. That reading misses what the five had in common. None of them was mainly about stopping an intruder at the door. Each one decided something about what would happen after someone got in: how much data would be there to take, whether it could be read, whether the entity could see who had touched it, and whether the software holding it had been built to resist abuse. Together they set the size of a breach rather than its likelihood.

This article argues that Part 500’s eighteen-month provisions were a set of loss-limiting controls, and that a covered entity’s ability to show them working decided how bad its next incident would be and how well it could explain it. The analysis falls into four parts. The first is what is kept: the disposal requirement that limits what is there to steal. The second is what is readable: encryption and the CISO-approved exception. The third is what is visible: user monitoring and the audit trail that let an entity see and reconstruct what happened. The fourth is what is built: the application security procedures that shape how the software holding the data behaves. A final section takes the objections.

The parts do not overlap. Disposal acts on the quantity of data. Encryption acts on its form. Monitoring and audit trails act on the record of access. Application security acts on the code through which access happens. A covered entity can be strong on one and weak on another, and an incident will expose whichever is weak.

The regulation defines a covered entity as any person operating under, or required to operate under, a license, registration, charter, certificate, permit, accreditation or similar authorization under New York’s Banking Law, Insurance Law or Financial Services Law. In our reading, institutions chartered under federal law, such as national banks and federal credit unions, are outside that definition unless they hold such a New York authorization. Some covered entities qualify for limited exemptions, and organizations should confirm with counsel which sections apply to them.

The deadline also had a reporting consequence. Section 500.17(b) required each covered entity to certify compliance with the regulation to the Superintendent by February 15 each year, covering the prior calendar year, and to keep the records, schedules and data supporting the certificate for five years. Where an entity had identified areas needing material improvement, it had to document them and the remedial work planned and underway. Once the eighteen-month provisions were in force, the evidence for each of them became part of what stood behind the February filing. A control that existed but left no record would be hard to certify.

What is kept: disposal as a limit on exposure

The first part concerns data that should no longer exist. Section 500.13 required policies and procedures for the secure disposal, on a periodic basis, of nonpublic information about individuals that was no longer necessary for business operations or other legitimate purposes.

Data that is not there cannot be taken

The logic is simple. A breach exposes what is stored. Records retained past their use add to the exposure without adding to the business. A customer file kept for a closed account, a scanned application from a declined loan, a claims attachment from a closed matter: each is a liability once its purpose has ended. Disposal reduces the size of every future breach at once, which no other control in the eighteen-month group does.

The scope of the duty

The disposal requirement applied to the two categories of nonpublic information that concern individuals. The first covered identifying information combined with data such as Social Security numbers, driver’s license numbers, account and card numbers, access codes and biometric records. The second covered health information created by or derived from a health care provider or an individual, a category that matters to insurers. Business information whose disclosure would harm the entity was outside the disposal duty, though nothing stopped an entity from applying the same discipline to it.

The two exceptions

Section 500.13 did not require disposal where other law required retention, or where targeted disposal was not reasonably feasible because of the way the information was maintained. Both exceptions are reasonable. Banks and insurers face record-keeping rules that require some records to be kept for set periods. And some systems, such as older core platforms and backup tapes, cannot easily delete one record without affecting others.

The exceptions also define the work. To rely on the first, an entity needs a retention schedule that maps each record type to the law that requires it to be kept. To rely on the second, it needs to know which systems cannot delete selectively, and why. Without those two documents, an entity cannot show that the data it still holds falls within an exception.

Where disposal breaks down

Disposal is hard because data spreads. A customer record lives in the core system, in a data warehouse, in reports exported to spreadsheets, in email attachments, in a vendor’s copy, and in backups of each. A policy that deletes the core record and leaves the copies has reduced little. The policy has to follow the data, which means knowing where the copies are. Disposal is an inventory problem before it is a deletion problem.

Backups and archives

Backups deserve their own attention. They exist to preserve data, which puts them in tension with disposal. An entity can reasonably keep backups for recovery for a defined period and let them expire. What it cannot easily defend is an archive that keeps every record indefinitely because no one decided otherwise. A retention period for backups, documented and enforced, is part of a disposal program.

Who decides what to keep

Retention decisions sit uneasily between functions. Business owners want data kept in case it is useful. Legal and compliance know which records the law requires. Security knows that every retained record adds exposure. No single function among them can make the decision alone. A workable program gives each record type an owner who approves its retention period, with legal confirming the minimum and security able to challenge anything longer. Without an owner, the default is to keep everything, and the disposal policy exists only on paper.

The owner also needs a reason to act. A business line that bears none of the cost of a breach has little incentive to delete. Reporting retained volumes by owner to senior management, alongside the exposure they represent, changes that. It turns a technical setting into a decision someone has to defend.

Copies held by vendors

Data held by service providers is part of the same problem. A vendor that processes customer records may keep its own copies, logs and backups. Part 500’s third-party service provider section, due by March 1, 2019, required covered entities to address the security of data held by those providers. For disposal, the contract is the tool: it should say how long the vendor may keep data, how it will return or destroy it at the end of the relationship, and how the entity will confirm that this happened. An entity that deletes its own copy and leaves the vendor’s in place has disposed of half the risk.

Measuring progress

A disposal program can be measured. Useful measures include the share of record types with an approved retention period, the share of systems with a working disposal method, the volume of records past their retention date, and the number of open exceptions. Each measure gives senior management a view of whether exposure is shrinking. Without measures, a program can run for years without anyone knowing whether it has reduced what an intruder would find.

Evidence a disposal program produces

A disposal program that works leaves evidence: a retention schedule approved by legal and compliance, a list of systems with their disposal method, logs or certificates of disposal runs, and exceptions recorded with reasons. Those records answer the examiner’s question, and they also answer the question after a breach of why certain data was still held. An entity that can show that the exposed data was within its retention schedule is in a better position than one that cannot say why it was there.

What is readable: encryption and the CISO’s exception

The second part concerns the form of the data that remains. Section 500.15 required controls, including encryption, to protect nonpublic information in transit over external networks and at rest.

What encryption changes

Encryption does not stop an intruder from reaching data. It changes what the intruder gets. Data that is encrypted, with keys held separately and well protected, is of little use without the keys. For a breach involving stolen media or copied files, encryption can be the difference between an exposure and an inconvenience. Its value depends on where the keys are and who can use them, which is why key management is part of any encryption control.

The exception and who signs it

The 2017 text allowed an entity that determined encryption was infeasible, in transit over external networks or at rest, to use effective alternative compensating controls reviewed and approved by its CISO. It required the CISO to review, at least annually, both the feasibility of encryption and the effectiveness of the compensating controls. That design put a named role on the hook for every exception. A gap in encryption was permitted only if the CISO had approved an alternative and kept reviewing it.

What made the exception reasonable

The exception reflected real constraints in 2018. Some older systems could not encrypt data at rest without replacement. Some data flows to counterparties relied on protocols that the entity did not control. Forcing encryption everywhere by a single date could have meant replacing core systems in months. The exception allowed an entity to protect the data by other means, such as network segmentation, strict access control and monitoring, while it planned the replacement.

What made the exception risky

An exception that is approved once and renewed each year without change becomes permanent in effect. The annual review requirement was designed to prevent that, but a review is only as good as its question. A CISO who asks each year whether encryption is still infeasible, and accepts the same answer, has complied with the letter of the rule while leaving the gap in place. The better question is what it would take to make encryption feasible, by when, and at what cost, with that plan presented to senior management.

Building the exception register

A covered entity that wanted to show control would keep an exception register: each system or data flow without encryption, the reason, the compensating controls, the CISO’s approval date, the date of the last annual review, and the plan to remove the exception. The register turns a scattered set of approvals into a view senior management and the board can see. It also gives the CISO a basis for the annual report to the board that section 500.04 already required.

Exceptions in front of the board

Section 500.04 already required the CISO to report in writing at least annually to the board or equivalent body on the cybersecurity program and material cybersecurity risks. An encryption exception is a material risk decision by any reasonable reading, because it leaves nonpublic information readable if taken. Including the exception register in that report puts the decision where the regulation expected material risks to go. It also lets the board ask the useful question each year: which exceptions were removed, which were added, and why.

Where unencrypted data turns up

An encryption program that covers databases and network links can still miss data in other places: laptops, removable media, file shares, exports to spreadsheets, email attachments and test environments loaded with copies of production data. Each is a place where nonpublic information can sit at rest outside the controls built for the main systems. The 2017 text applied to nonpublic information held or transmitted by the covered entity, which in our reading reaches copies wherever they sit. An entity mapping its encryption controls would start from where the data is, including the copies, rather than from where the encryption tools are installed.

Keys and who holds them

Encryption moves the problem from protecting data to protecting keys. If the keys sit on the same server as the data, or are available to every administrator, an intruder who reaches the server or an administrator’s account gets both. A sound design keeps keys in a separate system, limits who and what can use them, logs their use, and rotates them on a schedule. Those choices decide whether encryption reduces exposure or only satisfies a checklist. An examiner asking about section 500.15 can reasonably ask where the keys are and who can reach them.

Readability and the duty to notify

Whether data was readable can also affect what follows a breach. Notification duties under breach laws and contracts can turn on whether the exposed data could be used, and encryption with uncompromised keys is part of that analysis. The answer depends on each applicable law and each contract, and is a question for counsel in each incident. The practical point is that an entity that can show which data was encrypted, and that the keys were not taken, has more options than one that cannot.

Data in transit inside the network

The 2017 text required encryption in transit over external networks. In our reading, traffic inside the entity’s own network was left to the risk assessment. That was a judgment about where interception was likelier. It left internal traffic exposed to an intruder already inside, which is exactly the scenario the other eighteen-month controls address. An entity that relied on the external-only line needed its monitoring and segmentation to compensate.

What is visible: monitoring and the audit trail

The third part concerns the record of access. Two sections in the eighteen-month group dealt with it. Section 500.14(a) required risk-based controls to monitor the activity of authorized users and detect unauthorized access to, use of or tampering with nonpublic information. Section 500.06 required systems designed to reconstruct material financial transactions and audit trails designed to detect and respond to cybersecurity events.

Authorized users as the risk

Section 500.14(a) is about people who have legitimate access. That framing matters. An intruder who steals a valid credential becomes, to the systems, an authorized user. An employee who misuses access is one too. Controls that only watch for outsiders miss both. Monitoring authorized users means knowing what normal activity looks like for each role and flagging what departs from it: unusual volumes of record lookups, access at odd hours, access to records with no business link to the user’s work.

Risk-based, not total

The requirement was risk-based. An entity did not have to watch every action by every user with equal intensity. It had to decide, through its risk assessment, which users and which data warranted closer monitoring. Privileged administrators, staff with access to large volumes of customer records, and accounts used by vendors are natural candidates. The decision should be written down, because an examiner or an incident review will ask why some activity was monitored and other activity was not.

Two kinds of audit trail

Section 500.06 asked for two things that are easy to confuse. The first was the ability to reconstruct material financial transactions sufficient to support normal operations and obligations, with records kept for at least five years. That is about resilience: after an incident, can the entity rebuild its books and meet its obligations to customers? The second was audit trails designed to detect and respond to cybersecurity events that could materially harm normal operations, kept for at least three years. That is about investigation: can the entity see what happened and respond?

Why the retention periods matter

The retention periods show what the regulation expected investigations to need. Intrusions can go undetected for long periods. An entity that keeps security logs for weeks cannot investigate an intrusion that began months earlier. A three-year floor for detection records gave investigators a window measured in years rather than weeks. Keeping logs that long, in a form that can be searched, is a cost, and the regulation made it a requirement rather than a choice.

Logs that answer the questions

A log is useful only if it records the right things. For detection and response, the questions after an incident are who accessed what, when, from where, and what they did with it. Logs that record logins but not record-level access cannot answer whether a specific customer’s data was viewed. An entity reviewing its audit trails against section 500.06 would start from those questions and work back to the systems that must record the answers.

Testing reconstruction

The first audit-trail duty, the ability to reconstruct material financial transactions, is a resilience control, and resilience controls need testing. An entity can believe its records would let it rebuild its books after an incident and be wrong, because the records depend on the same systems the incident disabled. A test that asks a team to reconstruct a day’s transactions from the retained records, without the primary system, shows whether the duty is met. The test is good practice rather than an explicit requirement of the 2017 text, and it turns a stated capability into a demonstrated one.

Service accounts and vendors

Not every authorized user is a person. Service accounts used by applications, integration accounts used by vendors and scheduled jobs all access data with valid credentials. They can be harder to monitor because their normal activity is high volume and constant. They are also attractive to intruders for the same reason. A monitoring program that covers people and ignores these accounts leaves a large blind spot. Each such account should have an owner, a documented purpose, and a baseline of normal activity against which departures can be flagged.

Who reads the alerts

Monitoring produces alerts, and alerts need people. An entity can deploy tools that flag unusual access and still miss an intrusion if no one reviews the flags in time. The staffing decision, whether alerts are reviewed around the clock, during business hours, or by an outside provider, is part of the control. So is the decision about which alerts trigger immediate response and which wait. Those decisions belong in the risk assessment that section 500.14(a) relies on, written down with their reasons.

Visibility and notification

The record of access also bears on the regulation’s notice requirements. Section 500.17 required notice to the Superintendent within 72 hours of a determination that certain cybersecurity events had occurred. That determination depends on what the entity can see. An entity with poor monitoring and short logs may learn of an event late and struggle to say what it involved. One with good monitoring and long logs can determine scope faster and with more confidence. The eighteen-month controls fed directly into the quality of every later notice.

What is built: application security as a design control

The fourth part concerns the software through which data is accessed. Section 500.08 required written procedures, guidelines and standards to ensure the use of secure development practices for in-house developed applications, and procedures for evaluating, assessing or testing the security of externally developed applications used by the entity. The CISO or a qualified designee had to review and update them periodically.

Applications as the path to data

Customer data is reached through applications: online banking, policyholder portals, agent and broker systems, internal case management tools. A flaw in an application can expose data to anyone who finds it, without any need to defeat the network. Secure development practices aim to prevent those flaws from being built in. They cover how requirements are written, how code is reviewed and tested, how access controls are designed, and how changes are released.

Built in house and bought

The rule covered both kinds of software, with different duties. For applications the entity built, it required secure development practices. For applications it bought, it required procedures to evaluate, assess or test their security in the entity’s own environment. That second duty is easy to overlook. A purchased application configured by the entity, connected to its data and exposed to its users, is part of its attack surface whatever the vendor’s assurances. Testing it in place is how the entity learns what the configuration has done.

Access control as a design question

One damaging class of application flaw is the failure of authorization: a page that shows one customer’s document to anyone with the link, or an interface that returns any record if the caller changes an identifier. These flaws do not depend on sophisticated attacks. They depend on whether the application checks, for every request, that the requester is entitled to the data. Secure development procedures that require such checks, and tests that try to defeat them, address a class of exposure that network defenses cannot.

Evidence of secure development

An entity that wanted to show compliance with section 500.08 would keep its development standards, records of code review and security testing for releases, results of testing on purchased applications, and the CISO’s periodic review. Those records also matter after an incident. If a flaw is exploited, the question will be whether the entity’s procedures should have caught it, and the records are the answer.

Purchased software and the certification

Purchased applications raise a specific evidence question. The entity cannot see the vendor’s code, but it certifies its own compliance each year. The rule’s answer was to require procedures for evaluating, assessing or testing externally developed applications in the entity’s own environment. The evidence is therefore the entity’s own testing and review, supplemented by what the vendor provides. A file that holds only the vendor’s marketing assurances does not show that the entity evaluated anything.

Code the entity did not write

Applications built in house are assembled from components the entity did not write: open-source libraries, frameworks and vendor modules. A flaw in a component becomes a flaw in every application that uses it. Secure development procedures therefore need to cover how components are chosen, how their versions are tracked, and how quickly a known flaw in one is fixed. An entity that cannot list the components in its customer-facing applications cannot know which of them a newly announced flaw affects.

Changes under pressure

Secure development practices are tested hardest when a change is urgent. A production fix released at night to restore a service, or a feature pushed to meet a business date, is where review and testing are likeliest to be skipped. Procedures that allow emergency changes should also require review after the fact, with a record. Without that, the emergency path becomes the ordinary one for any team under pressure, and the written standard describes how software is meant to be built rather than how it is built.

Where application security meets the other parts

Application security connects to the other three parts. Applications that retain data longer than needed undercut disposal. Applications that store data unencrypted undercut encryption. Applications that do not log record-level access undercut monitoring. A development standard that addresses retention, encryption and logging as design requirements makes the other controls work. One that treats them as afterthoughts leaves each to be fixed separately.

Objections, and where the argument is weakest

The argument here is that the eighteen-month provisions were loss-limiting controls that set the size of a breach. There are fair objections.

The objection that prevention matters more

One could argue that the controls that keep intruders out, such as multifactor authentication and patching, matter more than those that limit loss once they are in. Prevention is essential, and Part 500 addressed it elsewhere. The objection does not weaken the argument. Prevention fails at some rate in every organization, and the eighteen-month controls govern what happens when it does. A program that invests only in prevention has no plan for its own failure.

The objection that the exception swallowed the rule

The encryption exception could be read as leaving encryption optional for anyone willing to sign an approval. That risk was real, and the later amendment discussed in the hindsight section narrowed it. In 2018 the exception still required a named approver, compensating controls that were effective, and an annual review. Used honestly, it was a transition device. Used as a permanent waiver, it defeated its own purpose. The difference lay in the entity’s governance, not in the text.

The objection that compensating controls can match encryption

A well-segmented network with tight access control and close monitoring can protect unencrypted data effectively, and in some environments it may be the better investment. The 2017 text accepted that, which is why it allowed the exception. The argument here does not claim encryption is always superior. It claims that the exception changed who carried the risk decision and that the decision needed to be revisited. Compensating controls protect data only while they keep working; encryption keeps protecting a copy after it has left the environment. The two are not equivalent once data is copied out, which is the case a breach creates.

The objection that risk-based requirements are too vague

Monitoring and audit trails were defined by reference to the entity’s risk assessment, which leaves room for judgment. One could argue that such requirements cannot be enforced or relied on. The flexibility is deliberate and has a cost: the entity must document its judgments. A risk-based requirement is enforceable when the entity’s written risk assessment and its controls do not match. The burden of showing they match falls on the entity.

The objection that small firms could not do this

The regulation’s limited exemption for small covered entities exempted them from several of the eighteen-month sections, including 500.06, 500.08, 500.14 and 500.15. It did not exempt them from 500.13 on disposal. The design suggests a judgment that disposal is within reach of any firm, while monitoring, audit trails, application security and encryption require more resources. Small firms that qualified still had good reason to apply the same thinking where they could.

The objection that this was one state’s rule

Part 500 is a New York regulation, and its terms bind only covered entities. A national bank, a federal credit union or a firm licensed elsewhere is outside it. One could argue that its design therefore matters only to New York licensees. The argument here does not depend on jurisdiction. The four effects described above, on what is kept, what is readable, what is visible and what is built, follow from how breaches work, not from where an entity is licensed. Any institution can ask the same four questions of its own program, whatever rules apply to it.

The objection that the deadline was arbitrary

The deadline was a date in a transition schedule, and nothing about September 4, 2018 made these controls more important than on any other day. The point is fair. Its importance was as a forcing device. After that date, an entity certifying compliance each February had to be able to certify these controls too, and the evidence to support the certification had to exist.

What we would tell a CISO today

Measure your exposure by what an intruder would find, as well as by whether one could get in. Keep a retention schedule and a disposal record that follow data to its copies. Keep an encryption exception register with a removal plan for each entry, and present it to the board. Make sure monitoring covers authorized users and that logs answer who accessed which record, kept long enough to investigate. And require your development and procurement processes to treat retention, encryption, authorization and logging as design requirements.

With hindsight

The regulation was amended. On November 1, 2023, the Governor announced that the Department had amended Part 500, in what the Department’s text calls the Second Amendment. The announcement described enhanced governance requirements, additional controls to prevent unauthorized access and the spread of an attack, more regular risk and vulnerability assessments, stronger incident response and continuity planning, and updated notification rules, including a new requirement to report ransomware payments.

Several of the eighteen-month provisions changed. In the amended text, section 500.15 requires a written policy requiring encryption that meets industry standards to protect nonpublic information both in transit over external networks and at rest. The compensating-control option now applies only where encryption of nonpublic information at rest is infeasible; such controls must be approved by the CISO in writing, and the CISO must review the feasibility of encryption and the effectiveness of the controls at least annually. The option the 2017 text allowed for data in transit over external networks is gone. The amendment gave covered entities one year from its effective date to comply with the new requirements in 500.15.

Section 500.13 became “Asset management and data retention requirements.” It now begins with a requirement for written policies and procedures to produce and maintain a complete, accurate and documented asset inventory, tracking for each asset, as applicable, its owner, location, classification or sensitivity, support expiration date and recovery time objectives. The disposal duty follows as a second subsection. The amendment gave covered entities two years from its effective date to meet the new requirements in 500.13(a). Section 500.14(a) now adds controls against malicious code, including web and email filtering, and annual awareness training that includes social engineering. Section 500.8 now requires the CISO or a qualified designee to review application security procedures at least annually.

The amendment repeated the staged approach of 2017. It gave covered entities 30 days from its effective date for the new requirements in section 500.17; one year for those in sections 500.4, 500.15, 500.16 and 500.19(a); 18 months for those in sections 500.5(a)(2), 500.7, 500.14(a)(2) and 500.14(b); and two years for those in sections 500.12 and 500.13(a).

Section 500.6 is one provision that kept its 2017 shape. The amended text still requires systems designed to reconstruct material financial transactions and audit trails designed to detect and respond to cybersecurity events, with the same five-year and three-year retention periods.

The annual filing changed too. The amended section 500.17(b) moves the filing date to April 15 and allows a covered entity either to certify that it materially complied during the prior year or to submit a written acknowledgment of noncompliance that identifies the sections not complied with and gives a remediation timeline, signed by the highest-ranking executive and the CISO.

What held up was the logic of the eighteen-month group. The amendment kept each control and tightened several, closing the transit exception and tying disposal to an asset inventory, which is the knowledge disposal depended on. What did not hold up was the idea that a signed exception could stand in for encryption of data moving over external networks. The amended text also requires the covered entity’s highest-ranking executive to sign the annual filing alongside the CISO, which makes the evidence behind these controls a matter for the top of the organization.

Spotted an error? Tell us and we will check it and correct the article where needed.

  • 23 NYCRR 500
  • NYDFS
  • encryption
  • data retention
  • audit trail