The SEC approved CAT in 2016 with an acknowledged breach risk, then approved taking investor identities out

The SEC’s November 15, 2016 order approved a central audit trail holding customer names, addresses and Social Security numbers, while calling the data an attractive target. Exemptions in 2020 and 2025 and a 2026 amendment removed those identity fields.

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

The SEC’s November 15, 2016 order approved a central audit trail holding customer names, addresses and Social Security numbers, while calling the data an attractive target. Exemptions in 2020 and 2025 and a 2026 amendment removed those identity fields.

  1. The week The plan was not the SEC’s own design.
  2. What the plan collected and why That last requirement is hard.
  3. How the plan said it would protect the data The plan’s security appendix set minimum requirements.
  4. What the Commission said it did not know It then explained why it approved anyway.
  5. Who carried the work Broker-dealers carried the reporting load.
  6. How CAT landed on different firms A large firm with its own order systems reported directly.
  7. Objections, in order The objection is that removing identity data slows investigations.
  8. Checks for anyone running a central store of personal data CAT is a regulatory system, but its design questions apply to any central analytic store that joins events to people.
  9. With hindsight CAT was built, and the identity question was reopened.

The week

On November 15, 2016, the Securities and Exchange Commission voted to approve a national market system plan to create the consolidated audit trail, known as CAT. [1] The SEC described CAT as a single, comprehensive database. [1] It would let regulators track all trading activity in the U.S. equity and options markets more efficiently and thoroughly. [1]

The plan was not the SEC’s own design. The national securities exchanges and the Financial Industry Regulatory Authority had submitted it jointly. [1] They did so under Rule 613 of Regulation NMS, which the SEC adopted in July 2012. [1] Rule 613 required those self-regulatory organizations to submit a plan for a consolidated audit trail capturing customer and order event information across all markets. [1] The SEC had voted to publish the plan for comment on April 27, 2016. [1]

The approval order, Release No. 34-79318, was dated the same day. [2] Its title was Joint Industry Plan; Order Approving the National Market System Plan Governing the Consolidated Audit Trail. [2]

The plan set out what would be reported. At each stage of an order’s life, from origination through routing, modification, cancellation and execution, broker-dealers and the self-regulatory organizations would submit details to a central repository. [1] Those details included a customer identifier, a broker-dealer identifier, the time of the event and the order’s symbol, price, size and type. [1] Data was generally due by 8:00 a.m. Eastern time on the day after the event. [1]

The plan also covered customers themselves. Under what it called the Customer Information Approach, each broker-dealer would assign a unique Firm Designated ID to each customer account. [2] Broker-dealers would submit an initial set of customer information to the central repository. [2] The order listed that information as including the customer’s name, address, date of birth, individual taxpayer identification number or Social Security number, and role in the account. [2]

Inside the repository, each customer would be identified by identifiers such as the taxpayer number or Social Security number and date of birth. [2] The plan processor would use them to map orders to specific customers across all broker-dealers. [2] Broker-dealers would send daily updates for new or changed accounts and identifiers. [2]

The SEC said it had changed the plan in response to comments. It strengthened several data security requirements, including for personally identifiable information. [1] It tightened clock synchronization for the self-regulatory organizations to within 100 microseconds of the time kept by the National Institute of Standards and Technology. [1] It expanded the advisory committee and moved up the deadline for proposals to retire systems that CAT would make obsolete. [1]

The fact sheet listed the security design. All CAT data would be encrypted at rest and in flight. [1] Data centers housing CAT systems would need certification by a qualified, unaffiliated third-party auditor. [1] The plan processor would have to comply with the NIST Cybersecurity Framework and designate a chief information security officer. [1]

The order itself was candid about risk. Responding to comments on aggregating confidential data in one place, the Commission said it agreed that CAT data would be a particularly attractive target for bad actors. [2] It said it believed the plan’s security requirements gave appropriate protection. [2] It also said it continued to believe there was significant uncertainty about the risk of a breach. [2]

The timetable came last. The self-regulatory organizations had two months to select a plan processor to build and operate CAT. [1] They would begin reporting within one year of approval, large broker-dealers the following year, and small broker-dealers the year after. [1]

The SEC framed CAT’s approval around market oversight: research, reconstruction of market events, monitoring and investigation of misconduct. [1] Those were the stated goals. For data and security leaders, the more instructive part of the November 15, 2016 order is narrower. It approved a central store of investor identities while saying, in writing, that the breach risk was significantly uncertain. [2]

Our argument is that the 2016 approval was a data minimization decision made in the other direction, with its risks stated openly. At CAT’s inception, as the SEC later put it, customer information was considered important to identify customers behind potentially unlawful trading more quickly and reliably. [4] In later years, the SEC revisited that choice step by step and approved taking the identity fields out, while keeping anonymized customer identifiers. [4] The path from one decision to the other shows how to decide what identity data a central analytic store should hold.

We take the 2016 decision in four parts:

  1. What the plan collected and why: the identity fields and the regulatory benefit claimed for them.
  2. How the plan said it would protect them: the security requirements written into the plan.
  3. What the Commission said it did not know: the attractive target, the uncertain breach risk and the unassigned breach cost.
  4. Who carried the work: the reporting duties that fell on broker-dealers and the self-regulatory organizations.

The first part is the purpose of the data, the second its protection, the third the residual risk the order admitted, and the fourth the operational burden. Together they cover the decision the order made. After them we look at how CAT landed on different kinds of firms, take the strongest objections in order, and end with checks for anyone running a central data store of personal information.

What the plan collected and why

The customer identifier problem

Rule 613 required a reporter to give a Customer-ID for each customer when it first received or originated an order. [2] The rule defined Customer-ID as a code that uniquely and consistently identifies a customer for reporting to the central repository. [2] It also required all plan sponsors and members to use the same Customer-ID for each customer. [2]

That last requirement is hard. Each broker-dealer knows its own customers by its own account numbers. No shared identifier existed across firms. A single customer trading through three firms appears as three unrelated accounts unless someone links them.

The self-regulatory organizations proposed a workaround. They asked for exemptive relief so the plan could use the Customer Information Approach. [2] The Commission granted that relief on March 1, 2016, so the approach could be put out for notice and comment. [2]

How the approach worked

Under the approach, each broker-dealer assigned a Firm Designated ID to each account. [2] The firm could use an account number or any other identifier, as long as it was unique within the firm for each business date. [2] Each new order carried only the Firm Designated ID. [2]

The link to the person came separately. Broker-dealers would submit identifying information for each customer to the central repository and keep it updated daily. [2] The plan processor would then use identifiers such as Social Security number and date of birth to tie each customer’s orders together across all firms. [2]

In data terms, CAT split into two stores from the start. One held order events keyed by firm-level identifiers. The other held identity data that resolved those identifiers to people. The SEC later described the design the same way: personal customer information was stored separately from transaction data, and the transaction database held only anonymized order and event data. [4]

Why one key mattered

The value of a shared customer key is cross-firm visibility. Rule 613’s requirement that all plan sponsors and members use the same Customer-ID for each customer is what makes that possible. [2] With it, a regulator can see one person’s orders across every broker-dealer they use. Without it, each firm’s view stops at its own walls.

The 2016 plan met the requirement by resolving firm identifiers to people inside the repository. [2] That choice put the data that makes the join work, and the data whose loss would hurt investors, into one central store. The later redesign kept the join and moved the hard data out. [4]

The benefit claimed

The SEC’s stated purpose was regulatory. Through CAT, regulators would have more timely access to trading data to conduct research, reconstruct market events, monitor market behavior, and identify and investigate misconduct. [1] The plan let regulators run complex queries and reconstruct market events and the status of order books at different times. [1]

Identity data served that purpose directly. A regulator who sees a pattern of suspicious orders wants to know who placed them. With names and addresses in the repository, the answer was a query. Without them, the regulator would have to ask each broker-dealer. The SEC’s January 2026 order said there can be a material difference between the time it takes to find a customer’s name in CAT and the time to obtain it from a broker-dealer. [4] It added that, in its staff’s experience, the extra time generally does not materially impede examinations and investigations. [4]

The plan did not treat every customer the same way. Where a validated Legal Entity Identifier was available for a customer, the plan said it could remove the need for other identifying information such as name, address or employer identification number. [2] The customer file could also carry a Large Trader ID where one applied. [2]

That detail shows a design choice the 2016 plan already made for institutions. For legal entities with a registry identifier, a standard code could stand in for raw identity data. For natural persons, no such public code existed, so the plan reached for Social Security numbers and dates of birth. [2] The later changes, in effect, built a substitute code for people.

The scale of what was proposed

The order made clear the repository would cover NMS securities, including options, and over-the-counter equity securities. [1] The plan anticipated that customer information could be limited to accounts with, or expected to have, reportable activity. [2] It said accounts that were open but had not traded in a given period might not need to be loaded until they traded. [2]

For a data leader, the scope is the point. This was identity data for customers across the U.S. equity and options markets, joined to their trading. We did not find a count of the people that meant in the order. The design reached every broker-dealer customer with reportable activity.

How the plan said it would protect the data

Encryption and storage

The plan’s security appendix set minimum requirements. All CAT data had to be encrypted in flight using industry standard practices such as TLS. [2] Personally identifiable information had to be encrypted at rest and in flight, including in archival storage such as tape backup. [2] Storing unencrypted PII was prohibited. [2]

The plan anticipated public cloud use. CAT data stored in a public cloud had to be encrypted at rest. [2] If cloud managed services would have access to the data, the key management arrangement had to be documented, including whether the provider or the plan processor held the keys. [2] Monitoring of when provider personnel could access or decrypt data, and a response plan for unauthorized access, also had to be documented. [2]

Data centers housing CAT systems had to be SOC 2 certified by an independent third-party auditor at least once a year. [2] CAT infrastructure could not be commingled with non-regulatory systems, and processing systems had to be segmented from others as far as feasible. [2]

Access to identity data

The plan treated PII as a separate class. By default, users entitled to query CAT data were not authorized to see PII. [2] The process for granting PII access had to be documented, and access followed a role-based, least-privilege model on a need-to-know basis. [2]

Multi-factor authentication was required for every login, including those able to reach PII, as the participants clarified. [2] A designated officer at each participant and at the Commission, such as the chief regulatory officer, had to certify at least annually that people with PII access were properly designated. [2] A full audit trail of PII access had to be kept, and daily reports listed everyone entitled to it. [2]

Query results could not include PII. Results, reports and bulk extracts had to display non-PII identifiers such as the Customer-ID or Firm Designated ID. [2] The PII behind those identifiers had to be gathered through a separate, restricted process. [2]

People and testing

The plan processor had to require that people with access to the repository use the data only for surveillance and regulation. [2] Everyone with access to the repository, except SEC employees and Commissioners, had to sign a Safeguard of Information Affidavit providing for personal liability for misuse. [2]

The plan required penetration testing and an application security code audit by a named third party before launch and periodically after. [2] The plan processor had to have a cyber incident response plan, approved by the operating committee. [2] It had to comply with the NIST Cybersecurity Framework, and the participants had to keep security protocols for their own handling of CAT data as rigorous as the repository’s. [1]

Standards by reference

The plan named external standards. They included NIST SP 800-53 Revision 4, NIST contingency planning guidance and NIST continuous monitoring guidance. [2] They also included FFIEC authentication guidance and ISO/IEC 27001. [2] The plan said the CAT system had to satisfy applicable database security regulations, including Regulation SCI. [2]

On paper, this was a demanding control set. It is a list a bank examiner would recognize. The question the order could not answer was how well it would be carried out, by a processor not yet chosen.

Recovery and resilience

Security in the plan also meant availability. The plan addressed business continuity and disaster recovery for the central repository. [2] Among its amendments, the Commission required the participants to give it an annual assessment of efforts to reduce the time needed to restore and recover CAT data at a back-up site. [2]

For a repository holding identity data, recovery has a security side. Every back-up copy is another place the data lives. The plan’s rule that archival storage of PII had to be encrypted applied there too. [2] A recovery test that restores data into an environment with weaker controls would undo the point of the primary controls.

Governance of the security officer

The order also dealt with a conflict of interest. The chief compliance officer and the chief information security officer would be employees of the plan processor. [2] Commenters raised the risk that their duties to the processor could conflict with their duties to CAT. [2]

The Commission modified the plan so both officers owed fiduciary duties to CAT LLC in the same manner and extent as an officer of a Delaware corporation. [2] Where those duties conflicted with duties to the processor, the duties to CAT LLC would control. [2] The Commission said the change should reduce uncertainty about whether security responsibilities would be carried out in a way that minimized breach risk. [2]

That is a useful pattern for any outsourced data platform. If the vendor employs the person responsible for security, the contract should say whose interest that person serves when the two diverge.

What the Commission said it did not know

The attractive target

Commenters had argued that the repository’s scale and scope would make it a particularly attractive target for well-funded hackers, individuals and nation-states. [2] They listed motives from theft to insider trading to market disruption. [2] The Commission agreed that CAT data would be a particularly attractive target for bad actors. [2]

It then explained why it approved anyway. It said it believed the extensive security requirements in the adopted plan gave appropriate and adequate protection. [2] It also noted that it had amended the plan to require background checks for participant staff using the system. [2] And it required an evaluation of the information security program against the highest industry standards for protecting data. [2]

The uncertainty

The order did not claim the risk was resolved. The Commission said there was a degree of uncertainty about the security measures the plan processor would actually implement, and so about the risk of a breach. [2] After its changes, it said the added information mitigated some of that uncertainty. [2] It continued to believe there was significant uncertainty about the risk of a breach. [2]

The order gave a reason for not saying more. The Commission recognized that publicly releasing too many details about security requirements could create additional risk. [2] That is a fair point about operational security. It also meant outside reviewers could not judge the controls in detail.

Who would pay for a breach

One commenter noted that the plan did not say who would be liable for a breach. [2] The Commission answered that the plan required the processor’s cyber incident response plan to address insurance for security breaches. [2] Details about who would bear the cost of a breach, and when, were left to later discussions on insurance and liability. [2]

The Commission was direct about the result. Because the response plan would not exist until a processor was chosen, it did not know whether or under what circumstances the processor would bear the cost of a breach. [2] It said this created some uncertainty about the processor’s incentives to minimize breach risk. [2] It approved the plan without further change on this point. [2]

Second-order effects

The order also weighed how people might react to the risk. In its earlier notice, the Commission had considered that service bureaus might stop offering CAT reporting if PII raised breach costs enough. [2] It had also considered that investors might move activity offshore or stop participating to keep sensitive data out of the repository. [2] The order recorded that the Commission had said, in that notice, that it did not believe the plan’s effect on breach risk or costs would be great enough to cause significant second-order effects. [2]

Reading the admission

In our reading, the order made a classic trade. It accepted an acknowledged, unquantified risk to investors’ personal data in exchange for a regulatory benefit it could describe. The admission matters because it was written down. A later Commission could revisit the trade against the same record, and it did.

For a CISO, the pattern is familiar from internal decisions. A business case names the benefit of holding sensitive data. The security review names controls. The residual risk and the breach cost owner are left for later. CAT’s order said so in public.

What an unquantified risk still allows

The order did not estimate the probability or cost of a breach. It did something else that is useful. It named the conditions under which the risk would be borne: an unchosen processor, an unwritten response plan and an undecided allocation of breach costs. [2]

Each of those conditions could later be tested. Once a processor existed, its controls could be assessed. Once the response plan existed, it could be read. Once costs were allocated, incentives could be judged. An approval that names its open conditions gives the next reviewer a checklist, even without a number.

Who carried the work

Broker-dealers

Broker-dealers carried the reporting load. They had to record order events contemporaneously and report them by 8:00 a.m. Eastern time on the following day. [1] Time stamps had to be at least as granular as the firm’s own order handling systems used. [1] The minimum was one millisecond, except for manual orders and allocation times, which allowed one second. [1] Business clocks had to be synchronized within 50 milliseconds of NIST time, with a one-second allowance for manual orders. [1]

Each firm had to assign and maintain Firm Designated IDs, submit customer information and send daily updates. [2] So every broker-dealer became a supplier of identity data to a regulatory system it did not control. Its own customer master file was now a source for a national database.

Error rates

The plan set an initial maximum error rate of five percent for reported data, subject to testing, adjustments at each launch and periodic review. [1] It described a phased approach to lowering the rate. [1] The plan processor would measure and report errors and provide correction tools. [1]

Under the plan’s appendix, errors would be sent to reporters on a T+1 basis. [2] Corrected data was due back by 8:00 a.m. Eastern time on T+3, and regulators were to have corrected and linked order and customer data by 8:00 a.m. on T+5. [2] For a firm, that is a standing data quality operation with a fixed correction deadline.

Reports owed to the SEC

The Commission added a series of reports. It required the participants to give it an annual evaluation of how error rates should apply by product, data element or other criteria. [2] It required an annual evaluation of how the processor and participants were monitoring error rates. [2]

It required a written assessment of clock synchronization standards within six months of the plan’s effectiveness, considering standards by type of industry member or system. [2] And it required a written report within 12 months on implementing coordinated surveillance among the self-regulatory organizations. [2]

Those reports turned the plan into an ongoing supervisory relationship. For the participants, each one was a deliverable with a date. For broker-dealers, each could lead to changed requirements, such as tighter clocks or lower error thresholds, after go-live.

The self-regulatory organizations

The exchanges and FINRA owned the plan. They would conduct CAT through a jointly owned, not-for-profit Delaware limited liability company. [1] An operating committee of all the self-regulatory organizations, each with one vote, would manage it. [1] An advisory committee including broker-dealers, institutional investors and a service bureau would give input. [1]

They also had to retire duplicative systems. The amended plan required them to file rule proposals to eliminate duplicative rules and systems within six months of approval. [1] Retirement would take effect once CAT data met minimum standards of accuracy and reliability. [1] Until then, firms could face reporting to old and new systems in parallel.

A detail on commercial use

One plan provision deserves attention. Participants had to limit use of CAT data obtained from the central repository to surveillance and regulatory purposes. [2] But a participant could use the raw data it reported to the repository for commercial or other purposes, if not prohibited by law. [2]

The distinction is between data a participant reported and data it pulled from the shared store. For a broker-dealer, it is a reminder that the regulatory channel and the commercial channel can carry the same records. In our reading, the plan’s barrier applied to the repository’s pooled data, not to a participant’s own submissions. [2]

The regulator as a user

The SEC itself would be a user of CAT. The fact sheet said a cross-divisional steering committee of senior staff was being formed. [1] It would design policies for Commission access to, use of and protection of CAT data, comparable to those for the self-regulatory organizations. [1] The order excluded SEC employees and Commissioners from the plan processor’s affidavit requirement. [2]

So the strongest access controls in the plan applied to the participants and the processor. The SEC’s own controls were to come from its separate policies. That is a normal arrangement for a regulator. It also meant the security of CAT depended on more than one organization’s controls.

How CAT landed on different firms

A large broker-dealer

A large firm with its own order systems reported directly. It had to build feeds for order events, clock synchronization, the Firm Designated ID scheme and the customer information file. [1] [2] Its timetable came one year after the self-regulatory organizations began reporting. [1]

The firm’s risk changed too. Its customer identity data now sat in two places: its own systems and a regulatory repository. A breach of either would expose its customers. Only one of those places was under the firm’s control.

A small broker-dealer using a service bureau

A small firm had a later start date. [1] It could also report through a vendor. The order recognized that service bureaus might reconsider offering CAT reporting if PII raised their breach exposure. [2] A small firm depended on that market staying open.

For such a firm, the contract with the service bureau became a security document. It needed terms on how customer identity data would be handled in transit to CAT, and on who would answer for a breach in the vendor’s systems.

A service bureau that serves several small firms is itself a concentration point. It holds identity data for all its clients’ customers on its way to CAT. The 2016 plan set requirements for the central repository and for participants’ handling of CAT data. [1] [2] A firm’s vendor sat outside both. Its controls depended on the firm’s contract.

An options market maker

Some firms gained relief from the plan’s design. The exemptive relief the plan reflected meant only options exchanges, not options market makers, would report information about market maker quotations. [1] For those firms, one large data feed stayed with the exchanges.

The example shows how the plan allocated work by who already held the data. Exchanges already saw the quotes. Broker-dealers already held their customers’ identities. The plan asked each to report what it had, and the identity data came from the firms closest to the customer. [1] [2]

A bank with a broker-dealer affiliate

CAT’s reporting duties fell on the self-regulatory organizations and broker-dealers. [1] In our reading, a bank holding company with a broker-dealer subsidiary would meet them through that subsidiary. CAT covered trading in equity and options securities, so the bank’s deposit and lending business sat outside it. [1]

For the holding company’s CISO, the effect was an enterprise data flow that ran through a regulated affiliate to an outside regulatory store. It belongs on the data inventory and the third-party risk map, even though the recipient is a regulator.

Advisers and wealth managers

In our reading, a registered investment adviser that is not a broker-dealer was not a CAT reporter under the plan’s structure. Its clients’ orders would still be reported, with identity data, by the broker-dealers that handled them. [2] The adviser’s clients’ data entered CAT through someone else’s report.

Objections, in order

1. Regulators needed identities in the repository to act fast

The objection is that removing identity data slows investigations. A regulator who must ask each broker-dealer for names loses time when markets are moving.

The SEC took this seriously when it later reconsidered. In 2026 it noted that removing foreign customers’ identifying information could make it harder to see the full extent of their trading and could delay swift, covert action. [4] It still approved the change, citing steps by the participants to reduce duplicate customer identifiers, a flag marking foreign customers’ trades and the customer records broker-dealers must keep. [4] It called that workaround less efficient and more time-consuming than current practice, but said it did not warrant disapproval. [4] The tension was real. The SEC resolved it by weighing it, not by denying it.

2. The security requirements were strong enough

The objection is that the plan’s controls were demanding, so the residual risk was acceptable.

The controls were demanding on paper. [2] But the 2016 Commission itself said significant uncertainty about breach risk remained, and that it did not know who would bear a breach’s cost. [2] Strong controls lower risk. They do not remove the consequences of holding the data if they fail.

3. Removing identities only moves the data back to broker-dealers

The objection is that the data still exists in firms’ systems, and regulators will still ask for it, so nothing is gained.

Something is gained: the data is no longer pooled in one place. Under the SEC’s approach, a regulator needing a customer’s identity would ask the reporting firm. [4] Commenters said those ad hoc requests increased firm costs. [4] Distributed data is not risk-free, but a single store of every customer’s identity is a different target from many separate firm systems.

4. A transformed Social Security number is still personal data

The objection is that a hashed or transformed identifier can still single out a person and, in some conditions, be linked back.

The SEC kept the transformation process as the way to create unique customer identifiers. [4] Under the 2025 exemption, reporters still sent a transformed value for the Social Security or taxpayer number. [3] In our reading, the change reduced what an intruder could take, but did not make CAT data anonymous in every sense. Data leaders should treat transformed identifiers as sensitive and protect the transformation method.

5. The original decision was right for its time

The objection is that hindsight is unfair. In 2016 the regulatory benefit was clear and the cloud security tools were less mature.

We agree the decision was defensible on its record. The order weighed the benefit and stated the risk. [2] Our point is narrower. A decision that states its residual risk openly can be revisited when the facts change. The SEC did exactly that.

6. The SEC should have waited for the processor’s security plan

The objection is that approving the plan before a processor, a response plan and a breach cost allocation existed put the decision ahead of the facts.

The order’s structure explains the sequence. Rule 613 required a plan first, and the plan set the processor selection process, with selection due within two months. [1] The Commission chose to approve a framework and require later evaluations and reports. [2] The cost was that the most important security facts arrived after approval. The benefit, from the SEC’s stated view, was a working audit trail on a defined timetable. [1]

Checks for anyone running a central store of personal data

CAT is a regulatory system, but its design questions apply to any central analytic store that joins events to people. These are the checks we would run. They are our suggestions, not requirements of the CAT plan or any rule.

  • Write down the specific decisions that need identity data in the store, and the time cost of getting it from the source instead.
  • Separate event data from identity data, and require a distinct, logged process to resolve one to the other.
  • Name the owner of breach costs before go-live, not after a processor or vendor is chosen.
  • State the residual risk in the approval record, so a later review can test it against what changed.
  • Treat transformed or tokenized identifiers as sensitive, and protect the method that produces them.
  • Set a review date for each identity field, with a default of removing fields no longer used.
  • Map every external store that receives your customers’ identity data, including regulators, onto your data inventory.

What the decision record should contain

CAT’s orders suggest the contents of a good approval record for any central store of personal data. We would include five items. First, the decisions that need identity data, stated as uses rather than as categories of data. Second, the controls relied on, with the open conditions still to be met.

Third, the residual risk, stated plainly even if it cannot be measured. Fourth, the owner of breach costs and the insurance position. Fifth, the trigger for review, such as a new threat assessment, a cheaper way to resolve identities or a change in how the data is used.

With those five items on file, a reversal like the one the SEC made becomes an ordinary governance step. Without them, removing data looks like an admission that the first decision was a mistake. The record changes the conversation.

Questions for a vendor that will hold identity data

The CAT structure placed a processor between the data’s owners and its users. Any firm that outsources a central store faces the same arrangement. Four questions follow from the CAT record.

  1. Who holds the encryption keys for identity data, and how is provider access to decrypted data monitored and logged?
  2. Whom does the vendor’s security officer answer to when the vendor’s interest and yours diverge?
  3. Who pays for a breach in the vendor’s environment, and is that written into the contract and the insurance?
  4. How would identity fields be deleted, and how would deletion be evidenced, if you decided to stop holding them?

The CAT plan answered the first two in 2016. [2] It left the third open. [2] The fourth became the subject of the 2026 amendment. [4]

The last check matters for broker-dealers in particular. A firm’s customer data in a regulator’s repository is outside its control. It is still data the firm supplied.

With hindsight

CAT was built, and the identity question was reopened. In March 2020, the SEC granted exemptive relief from the requirement to report Social Security or taxpayer numbers, dates of birth and account numbers. [3] The relief was conditioned on an alternative method of generating unique customer identifiers from transformed Social Security numbers. [3]

Under that method, a reporter transforms the Social Security or taxpayer number into an interim value. [4] That value, not the number itself, goes to a separate CCID subsystem, which generates the CAT customer identifier. [4] The 2020 relief limited the personal information reported to name, address and birth year. [4]

On February 10, 2025, the SEC went further. Its order granted exemptive relief from reporting names, addresses and years of birth for natural persons reported with transformed Social Security or taxpayer numbers. [3] The Commission said the creation of customer identifiers from transformed numbers had proven effective. [3] It said the benefits of continuing to collect those names, addresses and years of birth no longer justified the associated risks. [3]

The SEC explained its change of weighting later. It cited heightened security risks from increasingly sophisticated bad actors and the prospect of relatively efficient indirect access to customer information. [4] A regulator needing a person’s identity could use the firm identifiers tied to a customer identifier and ask the reporting broker-dealer. [4]

The plan itself changed next. On March 7, 2025, CAT LLC filed an amendment on behalf of the participants to reduce the customer information in the Customer and Account Information System. [4] On January 13, 2026, the SEC approved it, as modified by two amendments and by the Commission. [4] The order was published in the Federal Register on January 16, 2026. [4]

The approved amendment had broad effect. It eliminated all plan requirements to report names, addresses, years of birth, Social Security or taxpayer numbers, and employer identification numbers to CAT. [4] It required CAT LLC to direct the plan processor to delete previously reported customer data of that kind from the system, or otherwise make it inaccessible to regulatory users. [4] It codified the transformed-number method for generating customer identifiers. [4]

The participants said the change would save CAT LLC between $7 million and $9 million a year compared with the 2024 actual budget. [4] The SEC noted the cost to firms of answering more ad hoc requests for customer information. [4] It also noted that removing foreign customers’ identifying information could delay its efforts in some investigations. [4]

Funding became a second front. On July 25, 2025, the U.S. Court of Appeals for the Eleventh Circuit vacated the SEC’s 2023 order on CAT’s funding model and remanded it. [5] As the SEC later described the opinion, the court found the 2023 order arbitrary and capricious. [5] It said the SEC had not explained its decision not to bar participants from passing their share of CAT costs on to broker-dealers, and had not used actual CAT costs in its economic analysis. [5] On March 16, 2026, the SEC approved an amendment implementing a revised funding model, as modified by the Commission. [5]

That order also recorded the cost path. It cited CAT LLC’s projected 2026 budget of $156,432,998, compared with an initial estimated 2025 budget of $248,846,076. [5] It attributed the reduction to approved cost-savings measures, including the customer information amendment. [5]

The participants have said the identity side of the build ran late. In an exemption request of August 11, 2026, the participants stated that full implementation, including the customer and account system, was completed on July 15, 2024. [6] They said technical defects in that system had delayed the milestone past its December 30, 2022 target. [6] The SEC published notice of the request on September 1, 2026 and asked for comment. [6]

As checked on November 16, 2026, that request concerned the recovery of costs from the delayed period, and we did not find a Commission decision on it in the sources we opened. [6] In the sources we opened, we found no later change to the identity data requirements than the customer information amendment approved in January 2026. [4]

What held up from 2016 was the split design: order events in one store, identities in another, joined by identifiers. That design made the later changes possible without rebuilding the transaction database. [4] What did not hold was the judgment that the regulatory value of a central identity file outweighed its risk. The SEC reversed that judgment in stages, on the record. [3] [4]

In our view, the CAT history is a useful model for any organization reviewing its own data stores. The 2016 order recorded the risk it accepted. The 2025 and 2026 orders recorded why the balance changed. A decision trail like that makes reversal possible without pretending the first decision was wrong.

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

  • consolidated audit trail
  • SEC Rule 613
  • broker-dealers
  • PII
  • data minimization
  • FINRA