SHIELD’s safe harbor covered safeguards, not the state notice filings
New York’s SHIELD Act deemed banks and hospitals compliant with its security duty if they complied with the federal or state security rules they were already under. That harbor needed proof after a breach, and its notice duties had no comparable harbor.
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 SHIELD Act deemed banks and hospitals compliant with its security duty if they complied with the federal or state security rules they were already under. That harbor needed proof after a breach, and its notice duties had no comparable harbor.
- The week On July 22, 2019, the New York Legislature delivered Senate Bill S5575B to the governor.
- A harbor with a condition attached Section 899-bb gives an organization two ways to be “deemed” in compliance with its reasonable-security duty.
- The notice layer the harbor never reached The second half of SHIELD, the rewritten section 899-aa, has no deemed-compliance clause.
- A wider definition of breach Notice duties start with a breach, and SHIELD widened what counts as one.
- The perimeter outside the harbor The compliant regulated entity definition is about the entity, not the data.
- The file that makes the harbor hold If the harbor is a defense raised after a breach, the work that makes it hold is done before one.
- With hindsight Looking at New York’s law as it stood in 2026 confirms the split between the two halves of SHIELD.
The week
On July 22, 2019, the New York Legislature delivered Senate Bill S5575B to the governor. The bill’s formal name was the Stop Hacks and Improve Electronic Data Security Act, shortened to the SHIELD Act. According to the bill’s record on the New York Senate’s site, the governor signed it on July 25, 2019, as chapter 117 of the laws of that year. The bill had been introduced on May 7, 2019, at the request of the Attorney General, and both houses passed the final version in June.
The New York Attorney General’s office describes the act as an amendment to the state’s 2005 Information Security Breach and Notification Act. It did two things at once. It rewrote the breach notification section of the General Business Law, section 899-aa. And it added a new section, 899-bb, requiring any person or business that owns or licenses computerized data including New York residents’ private information to develop, implement and maintain reasonable safeguards for it.
The notification changes were broad. A breach no longer required that someone acquire data; unauthorized access to it was now enough. “Private information” grew to include biometric data used to authenticate a person, account and card numbers usable without any code, and a user name or email address combined with a password or security question and answer. The duty to notify now reached any person or business that owns or licenses New York residents’ private information. The old wording limited it to those that conduct business in the state; the bill struck it.
The bill also set out how the new law met the federal regimes banks and health systems already lived under. If an organization notified affected people under the breach rules of the Gramm-Leach-Bliley Act, HIPAA and the HITECH Act, or the Department of Financial Services’ cybersecurity regulation (23 NYCRR Part 500), it owed those people no second notice. It still had to notify the Attorney General, the Department of State and the State Police, and, when more than five thousand New York residents were to be notified at one time, the consumer reporting agencies. A HIPAA covered entity that reported a breach to the Secretary of Health and Human Services had to send that report to the Attorney General within five business days. That held even when the information involved was not “private information” under the state law.
Other changes tightened the edges. An organization could skip notice for an inadvertent disclosure by authorized people if it reasonably determined that misuse or harm was unlikely. It had to put that determination in writing and keep it for at least five years. If more than five hundred New Yorkers were affected, it had to give the determination to the Attorney General within ten days. For knowing or reckless violations, the civil penalty became the greater of $5,000 or up to $20 per instance of failed notification (up from $10), with the per-instance amount capped at $250,000 (up from $150,000).
The new section 899-bb set the security duty and its exits. An organization was deemed compliant if it was a “compliant regulated entity”: subject to, and in compliance with, the GLBA safeguards regulations, the HIPAA and HITECH rules, Part 500, or other federal or New York data security rules. Otherwise it could show a program with listed administrative, technical and physical safeguards, with a lighter standard for small businesses. Failing the duty was deemed a violation of General Business Law section 349, enforceable by the Attorney General. The section created no private right of action.
The bill made parallel changes to section 208 of the State Technology Law, which covers state entities. Most of the act took effect on the ninetieth day after it became law; the new security section took effect on the two hundred fortieth day.
New York’s SHIELD Act brought a reasonable-security duty to businesses that had no such state duty, while its sponsor’s memo promised no duplicate obligations for those already subject to other federal or New York data security rules. Banks, broker-dealers, insurers and health systems already answered to one or more such rules. For them, the headline duty to keep reasonable safeguards changed little on paper, because the act deemed them compliant if they complied with the federal or state rules they were already under.
That exemption is where the interesting part begins. The safe harbor was conditional, and on our reading the condition would be tested at the worst possible time: after a breach, by an Attorney General looking at what went wrong. Meanwhile the notice side of the law had no harbor at all. It added state filings, a broader trigger and new kinds of covered data on top of the federal clocks. Our argument is that SHIELD exempted regulated institutions from a duty they could already prove, and bound them to new duties that their federal programs did not cover.
The analysis falls into five parts, in the order an institution meets them:
- A harbor with a condition attached: what the deemed-compliance clause in section 899-bb actually gives a regulated entity, and what it asks in return.
- The notice layer the harbor never reached: the state filings and the HIPAA hand-off that apply whatever federal rule an institution follows.
- A wider definition of breach: how “access”, credentials and biometrics changed what counts as an incident to be assessed.
- The perimeter outside the harbor: the affiliates, vendors, practices and public bodies that never qualified as compliant regulated entities.
- The file that makes the harbor hold: the evidence an institution needs before an incident, so the exemption is still standing after one.
The parts do not overlap. The first two cover the two halves of the law, security and notice. The third covers the event that sets them off. The fourth covers who is in scope, and the fifth covers proof.
A harbor with a condition attached
Section 899-bb gives an organization two ways to be “deemed” in compliance with its reasonable-security duty. The first is to be a compliant regulated entity. The statute defines that as any person or business “subject to, and in compliance with” one of four sets of rules. They are the GLBA Title V regulations, the HIPAA and HITECH rules, the Department of Financial Services’ Part 500, or any other federal or New York data security rules as their own agencies or the courts interpret them.
The second way is open to everyone else. An organization builds a program with the administrative, technical and physical safeguards the statute lists. Those include a named coordinator, a risk assessment, training, vendor selection with contract terms, and regular testing of key controls. They also include disposal of data once it is no longer needed.
Two words carry the weight
The phrase “and in compliance with” is the whole design. Being a bank or a hospital does not make an organization a compliant regulated entity. Being in compliance with the rules that govern banks or hospitals does. The sponsor’s memo framed the clause as a safe harbor from enforcement for entities that take steps to verify their compliance with existing regulations. That wording points at verification, not status.
The consequence is easy to miss in a compliance calendar. The harbor does not shield a weak program. It points the state’s question back at the federal program and asks whether that program was actually being followed. A health system whose HIPAA risk analysis was stale, or a lender whose GLBA program skipped its vendors, may struggle to show it was “in compliance with” those rules. Once the harbor fails, the organization is measured against section 899-bb’s own list of safeguards, and because a failure is deemed a violation of section 349, the Attorney General can bring an action.
Why that matters more after a breach
The statute provides no filing to claim the harbor, and no agency certifies it in advance. On our reading it works as a defense, raised when the Attorney General investigates. That timing changes what an organization needs to have.
A supervisory examination reviews a bank’s information security program as a whole. An Attorney General’s inquiry after a breach would look at the program through the incident: which control failed, whether the risk assessment had named that risk, and whether the vendor had been held to safeguards by contract. The harbor holds only if the record answers those questions for the period before the breach.
So the clause shifted where the effort went. On this reading, SHIELD’s security section did not demand new controls of a regulated institution. It demanded that the existing program be demonstrable, control by control, at a moment the institution did not choose.
How an investigation would test the harbor
The statute does not prescribe how the Attorney General tests a claim of deemed compliance. The structure of section 899-bb suggests the sequence. The first question is whether the organization is subject to one of the listed regimes at all. The second is whether it was in compliance with that regime when the breach happened. The third, if it was not, is whether its program met the statute’s own list of safeguards.
The first question should be easy for a bank or hospital. The second is where the record decides the outcome. It asks about a period of time, not a point, and about the controls that failed, not the program in general. The third question gives a fallback, but only if the program’s documentation maps to the statute’s list, item by item.
We would expect an institution that has prepared for all three questions to answer quickly. One that prepared only for the first may spend the investigation reconstructing the second and third from scattered records, while the same people run the breach response.
The strongest objection
The strongest case against this reading is that the harbor was generous. Banks were already examined on their programs and health systems already answered to HHS under HIPAA, so, the argument runs, a state Attorney General would be unlikely to second-guess a program a federal or state regulator had accepted. On that view, SHIELD changed nothing for banks and hospitals, which fits the sponsor memo’s aim of avoiding duplicate obligations.
The objection is right about intent and about the ordinary case. It is weaker on the failure case, which is the case the law exists for. An examination is a sample. A breach is a specific failure with a specific cause. Later enforcement, covered under hindsight below, shows the Attorney General describing exactly those specific failures at health care providers: missing multi-factor authentication, accounts never decommissioned, a known vulnerability left unpatched. Those are questions a general examination rating does not settle.
What differs by type of institution
A New York-chartered bank that is a Part 500 covered entity can fall under two of the four listed regimes at once, the GLBA rules and Part 500. It can point to either, but it is safer when both programs tell the same story, because an investigator can ask about both.
A health system that is a HIPAA covered entity points to the HIPAA Security Rule. Its non-clinical lines, such as a fitness center, a parking operation, or a foundation’s donor database, may hold private information outside HIPAA’s scope. For those systems the harbor may simply not apply, a point the perimeter section takes up.
A broker-dealer or investment adviser points to whichever federal rules govern how it safeguards customer information. The fourth category, “any other data security rules” of a federal or New York agency, gives it room. It still needs the program to be one it actually follows. Which rule applies to which entity is a question for counsel. The statute only promises the harbor to entities that are subject to a rule and in compliance with it.
The notice layer the harbor never reached
The second half of SHIELD, the rewritten section 899-aa, has no deemed-compliance clause. It has a narrower accommodation, and the difference matters operationally.
What the accommodation covers
If an organization notifies affected people under the breach rules of GLBA, HIPAA and HITECH, or Part 500, section 899-aa says nothing in it requires additional notice to those people. That spares a bank or hospital a second, state-formatted letter to the same customers or patients.
The accommodation stops there. The same paragraph says notice “still shall be provided” to the Attorney General, the Department of State and the State Police. It also says notice still goes to the consumer reporting agencies when more than five thousand New York residents are notified at one time. The state filing must describe the timing, content and distribution of the notices and the approximate number of people affected. The 2019 text added a requirement to provide a copy of the template notice sent to affected people.
The HIPAA hand-off
For health systems, the provision written with them in mind is subdivision 9. A HIPAA covered entity that must notify the Secretary of Health and Human Services of a breach must also notify the New York Attorney General within five business days of notifying the Secretary. The statute applies this even to breaches of information that is not “private information” under the state definition.
That reaches further than it looks. HIPAA’s breach rule covers unsecured protected health information, which includes many records the 2019 state definition did not, such as diagnoses or treatment details without a Social Security number. Under subdivision 9, those breaches now produced a state filing too.
The timing follows the federal clock, not a fixed date. Under 45 CFR 164.408, a covered entity reports breaches affecting five hundred or more people to the Secretary at the same time as it notifies the individuals. It logs smaller breaches and reports them to the Secretary within sixty days after the end of each calendar year. The New York filing is due within five business days of whichever federal report is made. A health system therefore needs the state step built into both its incident playbook and its annual small-breach submission.
Two clocks that do not start together
HIPAA’s clock runs from discovery. Under 45 CFR 164.404, a breach is treated as discovered on the first day it is known to the covered entity, or would have been known with reasonable diligence. Notice is due without unreasonable delay and no later than sixty calendar days after discovery.
The 2019 New York text used a different measure: the most expedient time possible and without unreasonable delay, allowing for law enforcement needs and for measures needed to determine the scope of the breach and restore the system’s integrity. Both standards turn on “unreasonable delay”, but they were written by different authorities and would be judged by different ones. A covered entity that waited until late in HIPAA’s sixty days could find the state asking whether that was the most expedient time possible. Hindsight shows the state later replaced its standard with a fixed number.
What the notice has to say
SHIELD also set what a notice to residents must contain, whatever method is used. It must give contact information for the organization making the notice. It must give the telephone numbers and websites of the relevant state and federal agencies that provide information on breach response and identity theft prevention. And it must describe the categories of information accessed or acquired, specifying which elements of personal and private information were involved.
For an organization that notifies under HIPAA or GLBA, the accommodation means the federal letter is the letter. But the state filing must include a copy of the template that went out. A template drafted only to federal content rules will be read by a state office that knows its own content rules. We would align the two in advance.
Substitute notice at scale
Large institutions should know the substitute notice route before they need it. An organization may use it if it shows the Attorney General that direct notice would cost more than $250,000, that more than 500,000 people would need notice, or that it lacks sufficient contact information. Substitute notice then means all of three things: email where the organization has addresses, conspicuous posting on its website, and notice to major statewide media.
The credential rule cuts across this. If the breached data includes email addresses with passwords or security answers, email notice is not allowed; the likely reason is that the attacker may control the inbox. The organization must instead show the notice online to people connecting from their usual locations. A bank whose online banking credentials were exposed would need its digital channel ready to carry that notice, which is a product and engineering task, not a letter.
Law enforcement delay
Notice may be delayed if a law enforcement agency determines that it would impede a criminal investigation. It must be made once the agency determines that notice will not compromise the investigation. The statute ties the delay to the agency’s determination, not the organization’s preference. A record of who in law enforcement asked for the delay, when, and when they released it belongs in the incident file.
Who owns the filings
Inside an institution, the federal notices can sit with an existing owner: a privacy officer for HIPAA, a compliance function for GLBA or Part 500. In our judgment the state filings are easy to orphan because they are not part of any federal program. Our suggestion, not a statutory requirement, is to name one owner for every state filing in the incident plan, with the template notice and the count of New York residents as standing inputs.
Counting New York residents is its own task. It requires reliable address data for every affected person, which an institution may not have for prospects, former customers or people whose records came from a vendor. The consumer reporting agency threshold of five thousand residents “at one time” turns on that count. A bank that cannot produce the count quickly cannot know whether it owes notice to the consumer reporting agencies.
A wider definition of breach
Notice duties start with a breach, and SHIELD widened what counts as one. Three changes did the work, and each lands differently in a regulated institution.
Access as well as acquisition
The 2005 law turned on unauthorized acquisition of computerized data. SHIELD added unauthorized access. It also told organizations what they may consider in deciding whether information was accessed: indications that it was “viewed, communicated with, used, or altered” by a person without valid authorization.
For incident response, this moves the question from exfiltration to exposure. Under an acquisition standard, an institution could focus on whether data left the network. Under an access standard, an intruder who opened a file share and browsed records may have caused a reportable breach even with no evidence of copying.
That raises the value of a particular kind of log. File access logs, database query logs and application audit trails become the evidence that decides whether notice is owed. An institution that cannot show what an intruder viewed may be unable to rule out access. Without that, the safer reading may be to notify.
The definition keeps one carve-out. Good-faith access by an employee or agent for the organization’s purposes is not a breach, provided the information is not used or subject to unauthorized disclosure. That puts weight on role-based access. An employee browsing records outside their job is not acting for the organization’s purposes, and only access controls and audit trails can show which side of the line a given look fell on.
Credentials as private information
SHIELD made a user name or email address combined with a password or security question and answer that would permit access to an online account into private information on its own. No Social Security or account number is needed.
For a bank, that brings online banking credential theft into the state breach law directly. It also changed how notice may be given. Where the breached information includes such credentials, substitute notice by email is not allowed. The organization must instead give clear and conspicuous notice online, when the person connects from a location the organization knows the person customarily uses.
For a public agency or a health system with a patient portal, the same change applies to portal accounts. A credential-stuffing incident against a portal, in which attackers log in with passwords stolen elsewhere, now needs a fresh analysis under the state law. That holds even when no medical record is touched.
Encryption, and the key that goes with it
Since 2005 the definition of private information has excluded encrypted data unless the encryption key was also acquired. SHIELD widened that exception by adding access. Data now counts as private information if it is not encrypted, or if it is encrypted with a key that has also been accessed or acquired. A key that an intruder viewed, without any sign of copying it, may be enough.
That makes key management part of breach analysis. Encryption at rest helps only if the keys sit somewhere the intruder did not reach. A database encrypted with keys stored on the same server, or reachable by the same compromised administrator account, may give no protection under the statute.
For a bank or health system, the operational question is where keys live and who can use them. A hardware security module or a cloud key management service with separate administrative roles gives a stronger answer than keys stored with the data. The answer has to be provable after an incident, which means key access logs as well as key storage.
Biometrics and account numbers
SHIELD also added biometric information used to authenticate or identify a person, such as a fingerprint, voice print, or retina or iris image. It added account, credit or debit card numbers that could be used to access a financial account without any additional code or password.
Voice authentication in a call center and fingerprint time clocks in a hospital both produce biometric data in this sense. Neither may be managed as part of the customer data program. When the definition changed, the inventory of where private information lives had to change with it.
The inadvertent-disclosure exception and its paperwork
The new exception for inadvertent disclosure by authorized people is narrower than it looks. It requires a reasonable determination that misuse, financial harm, or emotional harm in the case of online credentials is not likely. The determination has to be in writing and kept for at least five years. For incidents affecting more than five hundred New Yorkers, it has to go to the Attorney General within ten days.
The exception rewards institutions that can reason about an incident quickly and write that reasoning down. A misdirected email with a spreadsheet of customer data is one example. To use the exception, someone has to decide, document and, above the threshold, file. Without a template and an owner, the exception is hard to use in time.
The perimeter outside the harbor
The compliant regulated entity definition is about the entity, not the data. Every legal person in a corporate group, and every vendor holding data for it, has to qualify on its own. That leaves a perimeter of organizations around banks and health systems for which the harbor may not exist.
Affiliates and side businesses
A bank holding company may own subsidiaries that hold New York residents’ private information but are not themselves under the bank’s safeguards rules: a marketing affiliate, a payroll unit or a real estate arm. A health system may run businesses that hold private information outside HIPAA’s reach.
For those entities, the route to deemed compliance is the second one: a program with the statute’s listed safeguards. The listed safeguards are close to what a bank or hospital program already contains. Extending the parent’s program to the affiliate, and recording that it was extended, is likely the cheapest path. What fails is assuming the parent’s status covers the child.
Vendors
SHIELD reaches vendors in two ways. First, section 899-aa requires an organization that maintains private information it does not own to notify the owner or licensee of a breach. The 2019 text said this must happen immediately following discovery. Second, the statute’s list of administrative safeguards includes selecting service providers capable of maintaining appropriate safeguards and requiring those safeguards by contract.
For a vendor that is itself a HIPAA business associate, the harbor may be available through the HIPAA rules. Other vendors, such as a mailing house, a marketing analytics firm or a collection agency, may have to show their own program. For the institution, the practical effect is contractual. Vendor agreements need a notice clause that matches the state duty, a description of the safeguards the vendor keeps, and a right to evidence that they are kept.
Small practices and small businesses
The statute defines a small business by any one of three tests. They are fewer than fifty employees; less than three million dollars in gross annual revenue in each of the last three fiscal years; or less than five million dollars in year-end total assets. A small business complies with the program route if its safeguards are appropriate to its size, complexity, activities and the sensitivity of the data.
A small physician practice or dental office that is a HIPAA covered entity can claim the regulated-entity harbor if it complies with HIPAA. If it does not, the small-business flexibility applies to the program route, not to the notice duties. The notice duties and the five-business-day HIPAA hand-off apply regardless of size.
State entities
Public agencies sit under a separate regime. The act amended section 208 of the State Technology Law for state entities. Its private information definition matched the business one, and it carried the same notice accommodation for agencies notifying under GLBA, HIPAA or Part 500. A state entity notifies the Attorney General, the Department of State and the state Office of Information Technology Services, rather than the State Police.
The state entity provisions carry things the business side lacks. A state entity already had to consult the Office of Information Technology Services on the scope of the breach and restoration measures. SHIELD added that, within ninety days of the notice of the breach, that office must deliver a report to the entity on the scope of the breach and recommendations to restore and improve the system’s security. It also required the office to develop, update and provide regular training to all state entities on preventing breaches.
For an agency program director, this creates a dependency outside the agency. The scoping and the post-breach report run through a central office. A state health or benefits agency planning its incident response needs that office in the plan from the first hour, not after the notices go out.
How long the exposure lasts
SHIELD also lengthened the period in which the Attorney General can act on a notification failure. An action under section 899-aa must be commenced within three years after the Attorney General became aware of the violation, or after the date of the state notice, whichever is first. No action may be brought more than six years after the organization discovered the breach, unless it took steps to hide the breach.
For the perimeter, this matters because affiliates and vendors may keep records for shorter periods than the parent. An affiliate that deleted its incident records after two years could face questions about a breach it can no longer document. Record retention for incident files should run at least as long as the statute’s exposure, and the five-year retention for written determinations sets a floor.
The file that makes the harbor hold
If the harbor is a defense raised after a breach, the work that makes it hold is done before one. The useful output is a file an institution can hand over quickly, which shows both that it was subject to a listed regime and that it was in compliance with it. None of what follows is a statutory checklist; it is what the statute’s wording suggests an investigator would ask for.
A map from the statute to the program
The first item is a crosswalk from the safeguards listed in section 899-bb to the institution’s existing program. Each listed safeguard gets the control that meets it, the policy that requires it and the last evidence that it operated. The list includes the coordinator, the risk assessment, training, vendor oversight, network and software risk assessment, detection and response, testing, physical protections and disposal.
The crosswalk serves two purposes. It shows the regulated-entity route, by showing the federal program in operation. And if the harbor is contested, it shows the program route as a fallback, because the same controls meet the statute’s own list.
Proof that the program was followed
Policies show intent; the harbor needs operation. The evidence that matters is dated: the most recent risk assessment and the remediation tracked from it, access reviews, patch records for critical systems, vendor reviews and training completion. Each should be tied to the period before any incident.
Two later New York settlements, described below, show where the Attorney General looked: whether multi-factor authentication was in place, whether inactive accounts were removed, whether credentials were rotated, whether a known vulnerability was patched. An institution that can produce those records by system is in a stronger position than one that can produce only an annual attestation.
The notice machinery
The file should also hold the notice side, since the harbor does not cover it. That means the state notice templates, the list of recipients for each regime, and the method for counting affected New York residents. For a covered entity, it means the five-business-day step tied to both the immediate and the annual HIPAA submissions. For any organization using the inadvertent-disclosure exception, it means the written determinations, kept for five years.
Testing the file before it is needed
A tabletop exercise is the cheapest way to find gaps in the file. We would run one that starts after containment, when an investigator asks for the program evidence. The scenario can be ordinary, such as a stolen administrator credential used to browse a file share holding customer records.
The exercise asks the questions the statute implies, and each unanswered one is a gap to close:
- Who decides whether the browsing was access to private information?
- Which logs show what was viewed, and for how long are they kept?
- Was the file share inside the current risk assessment?
- Were the encryption keys out of the intruder’s reach, and can the logs show it?
- Who files with the Attorney General, the Department of State and the State Police, and with what count of New York residents?
Measures a board can read
Boards and audit committees do not need the crosswalk. They need a few measures showing the harbor is likely to hold. Examples are the share of systems holding private information that appear in the current risk assessment, and the share of vendors holding such data with contract terms and a current review. Others are the time to produce a count of affected New York residents in the last exercise, and whether every state filing has a named owner.
Those measures are our suggestion, not a requirement. Their value is that each one ties to a question an investigator would ask, so movement in them means movement in the institution’s real position.
What it costs
For a bank or health system with a mature program, we expect most of this to be assembly rather than new control work. The cost sits in three places. First, keeping evidence tied to dates and systems, which needs tooling and discipline. Second, extending the program to affiliates and side businesses that were never in scope. Third, rewriting vendor contracts to carry the state notice and safeguard terms.
The trade-off is real. Every hour spent documenting a control is an hour not spent improving it. The answer is not to document everything. It is to document the controls an investigator would examine after the most likely breaches: remote access, identity, patching, logging and vendors.
SHIELD’s security duty asked regulated institutions mainly to prove what they already claimed. Its notice duties asked them to do new things. An institution that treated the law as “already covered” was right about the first half and exposed on the second.
With hindsight
Looking at New York’s law as it stood in 2026 confirms the split between the two halves of SHIELD. The security section has kept its structure. An Internet Archive copy of section 899-bb captured in August 2026 still defines a compliant regulated entity as one “subject to, and in compliance with” the GLBA, HIPAA and HITECH, or Part 500 rules. It still offers the program route with the same listed safeguards, and it still creates no private right of action.
The notice section kept moving. On December 21, 2024, the governor signed Senate Bill S2659B as chapter 647 of the laws of 2024, according to the bill’s record on the Senate’s site. The bill’s summary says it requires notice of a breach within thirty days and adds the Department of Financial Services to the agencies that must be notified. The archived text of section 899-aa from August 2026 reflects both changes.
Disclosure to affected residents is still to be made in the most expedient time possible and without unreasonable delay. It is now also bounded: notice “shall be made within thirty days after the breach has been discovered,” except for the legitimate needs of law enforcement. The 2019 allowance for time to determine the scope of the breach and restore the system no longer appears in that sentence. An organization holding data it does not own must still notify the owner immediately, and now within thirty days following discovery. Notice to the Department of Financial Services is required only of Part 500 covered entities, and is to be given in compliance with section 500.17.
The archived 2026 text also lists two more kinds of private information than the 2019 law did. They are medical information, covering a person’s medical history, condition, treatment or diagnosis by a health care professional, and health insurance information, such as a policy or subscriber number and claims history. The five-business-day hand-off for HIPAA covered entities remains in subdivision 9.
For health systems, those changes sharpened the timing question. HIPAA’s federal outer limit, as the eCFR shows in 2026, is still sixty calendar days after discovery. New York’s statute now names thirty days. The statute still says notice made under the HIPAA rules requires no additional notice to the same people. How those provisions fit together for a particular breach is a question to settle with counsel before an incident, not during one.
Enforcement showed what an investigation looks at. In January 2024, the Attorney General announced an agreement with Refuah Health Center over a May 2021 ransomware attack. The release says the attack compromised the information of approximately 250,000 New Yorkers. It says the office found the provider had failed to decommission inactive accounts, rotate credentials, limit staff access, use multi-factor authentication and encrypt patient information. The agreement required a $1.2 million investment in security and $450,000 in penalties and costs.
In December 2024, the office announced a $550,000 agreement with HealthAlliance. That followed an attack between September and October 2023 through a vendor-flagged vulnerability the operator knew of but had not patched. The release puts the affected patients at 242,641. Among other terms, it requires a policy of patching critical vulnerabilities within 72 hours or neutralizing them.
Those releases describe failures in specific, ordinary controls. That is the level at which a deemed-compliance claim would be tested. The 2019 reading that the harbor rewards proof, not status, still fits the unchanged wording of section 899-bb. What changed is the notice side, which grew a fixed clock and new data types. In our view, that makes it the part of SHIELD most likely to catch a regulated institution that thought it was covered.
Spotted an error? Tell us and we will check it and correct the article where needed.