The SEC’s 2017 cyber sweep found written policies at nearly every firm examined, and gaps in following them

On August 7, 2017, SEC examination staff reported on 75 broker-dealers, advisers and funds. Written policies were close to universal; the gaps were in tailoring, adherence, patching and, for advisers and funds, incident planning. In our reading, it made a firm’s own records the test.

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

On August 7, 2017, SEC examination staff reported on 75 broker-dealers, advisers and funds. Written policies were close to universal; the gaps were in tailoring, adherence, patching and, for advisers and funds, incident planning. In our reading, it made a firm’s own records the test.

  1. The week On Monday, August 7, 2017, the Securities and Exchange Commission’s Office of Compliance Inspections and Examinations, known as OCIE, published a National Exam Program Risk Alert titled Observations from Cybersecurity Examinations [1].
  2. The wording gap: policies too general to follow The first issue the staff listed was tailoring.
  3. The adherence gap: policies the firm did not follow, or practices it never wrote down The second issue was adherence.
  4. The maintenance gap: patches, unsupported systems and test findings left open The third part concerns system upkeep.
  5. The response gap: plans for incidents and customer notice that some firms lacked The fourth part concerns what happens when controls fail.
  6. How the alert reached different firms The alert covered three kinds of SEC registrant.
  7. What an evidence test leaves out The alert’s method has limits, and firms should know them.
  8. Objections, in order The alert said so itself.
  9. What a chief compliance officer could check The alert’s method can be applied internally.
  10. With hindsight Later SEC actions moved several of the alert’s concerns from staff observation into rule text, and dropped others.

The week

On Monday, August 7, 2017, the Securities and Exchange Commission’s Office of Compliance Inspections and Examinations, known as OCIE, published a National Exam Program Risk Alert titled Observations from Cybersecurity Examinations [1]. It was Volume VI, Issue 5 of the office’s risk alert series. The alert summarized what examination staff had seen in a cybersecurity sweep of firms registered with the SEC.

The sweep was called the Cybersecurity 2 Initiative. OCIE had announced it on September 15, 2015, in an earlier risk alert. Staff examined 75 firms: broker-dealers, investment advisers and registered investment companies, which the alert called funds. A footnote said the examinations ran from September 2015 to June 2016. They generally covered a review period from October 1, 2014 through September 30, 2015.

The alert placed the sweep in a sequence. OCIE had run a Cybersecurity 1 Initiative in 2014 and published a sweep summary on February 3, 2015. The second sweep looked at a different population of firms. The alert said it involved more validation and testing of procedures and controls than the first. A few of its observations had already appeared in a ransomware risk alert dated May 17, 2017.

The staff described its method in one sentence that carries the alert’s weight. The examinations focused on firms’ written cybersecurity policies and procedures, including validating and testing that they were implemented and followed. Staff also looked at six areas: governance and risk assessment, access rights and controls, data loss prevention, vendor management, training, and incident response.

The alert reported improvement since 2014. All broker-dealers, all funds and nearly all advisers examined kept written cybersecurity policies on protecting customer records and information. The staff said comparatively fewer broker-dealers and advisers had such policies in the first sweep. Nearly all broker-dealers ran periodic risk assessments of critical systems. All firms used some tool to prevent, detect and monitor loss of personally identifiable information.

It then reported where the picture was weaker. Nearly all broker-dealers ran penetration tests and vulnerability scans on critical systems, but the staff counted almost half of advisers and funds doing so. A number of firms did not appear to fully remediate some high-risk findings from those tests. A few firms had a significant number of uninstalled patches that, according to the firms, included critical security updates.

The alert’s section on issues opened plainly. The staff observed one or more issues in the vast majority of the examinations. A majority of firms’ information protection policies appeared to have issues, even though nearly every firm had such policies. Some were not reasonably tailored. Others were not followed, or did not describe what the firm actually did.

The alert tied these findings to two SEC rules. Regulation S-P, in 17 CFR part 248, subpart A, covers the privacy of consumer financial information and the safeguarding of customer records. Regulation S-ID, in subpart C, covers identity theft red flags. The staff described Regulation S-P-related issues among firms that did not appear to adequately maintain their systems.

The alert also listed elements the staff had seen in the policies of firms it believed had implemented strong controls. They included inventories of data and vendors, detailed instructions, prescriptive testing schedules, enforced access controls, mandatory training with completion tracking, and senior management approval of policies.

The document carried the usual staff caveats. It said the views were those of OCIE staff and that the Commission had expressed no view on its contents. It was not legal advice. The factors it described were not exhaustive and would not constitute a safe harbor. Its conclusion said cybersecurity remained one of the top compliance risks for financial firms, citing OCIE’s 2017 examination priorities of January 12, 2017.

A risk alert is a report of what examiners saw. It creates no new rule. The August 2017 alert still mattered because of what it chose to measure. The staff said this sweep involved more validation and testing of procedures and controls than earlier examinations. It asked whether the policies matched what firms did, and the staff said it tested that directly [1].

The answer was uneven. Nearly every firm had written policies on protecting customer information. A majority of those policies appeared to have issues. The gap was between paper and practice, and the alert described it in specific terms: vague instructions, reviews done less frequently than promised, training required but not enforced, patches not installed, test findings not fixed.

Our argument is that the alert moved the examination question from whether a policy exists to whether a firm’s own records show the policy works. Everything examiners criticized could be found by comparing a document with evidence. We take the gaps the staff described in four parts:

  1. The wording gap: policies too general to follow.
  2. The adherence gap: policies the firm did not follow, or practices it never wrote down.
  3. The maintenance gap: patches, unsupported systems and test findings left open.
  4. The response gap: plans for incidents and customer notice that some firms lacked.

The first part concerns what the policy says. The second concerns whether people did what it says. The third concerns whether systems were kept in the state the policy assumes. The fourth concerns whether the firm had planned for the moment those controls fail. Together they cover the distance between a written program and its operation. After the four parts we look at how the alert reached different firms, at what an evidence test leaves out, at objections, and at what a chief compliance officer could check.

A note on sources and our reading. We describe the 2017 findings from the alert itself. We describe later rules and priorities from the SEC’s Federal Register releases and its examination priorities documents. Statements about what a rule requires of a given firm are our reading. Firms should confirm their obligations with counsel.

The wording gap: policies too general to follow

The first issue the staff listed was tailoring. Policies were not reasonably tailored, the alert said, when they gave employees only general guidance or offered limited examples of safeguards. Others were very narrowly scoped, or vague because they did not set out procedures for carrying out the policy [1].

The phrase “reasonably tailored” has a regulatory echo. Regulation S-P requires reasonably designed policies and procedures with administrative, technical and physical safeguards for customer information, as the SEC’s examination staff later restated [5]. In our reading, the staff was signaling that a policy copied from a template could fail that standard even if it covered every topic. A policy that could belong to any firm says little about how this firm protects its data.

What a vague policy looks like

The alert gave categories, not quotations, so the following is our illustration. A policy might say only that the firm restricts access to sensitive data to authorized personnel. That sentence names no system, no approver, no review interval and no evidence. An examiner cannot test it, because nothing in it can be checked against a record.

A tailored version names the systems that hold customer data. It says who approves access to each, how requests are logged, how frequently access is reviewed, and what happens when someone changes role. The alert’s list of stronger practices described exactly this kind of detail for access rights. Requests were tracked, and procedures addressed onboarding, role changes and terminations.

Why tailoring starts with an inventory

The first element on the staff’s list of stronger practices was an inventory. Policies at those firms included a complete inventory of data and information, risk classifications, and information on each service provider and vendor. This element comes first for a reason. A firm cannot tailor a policy to systems it has not listed.

Inventory work is unglamorous and hard to finish. Data sits in core platforms, in shared drives, in mailboxes, in vendor systems and in spreadsheets copied for one-off analysis. A broker-dealer may hold customer data in its clearing firm’s systems as well as its own. An adviser may hold it in a portfolio accounting platform, a CRM and a document vault run by three vendors.

The inventory also defines the scope of everything else. Patching applies to listed systems. Access reviews apply to listed systems. Vendor oversight applies to listed vendors. Any gap in the inventory becomes a gap in each control that relies on it. For that reason, in our view, the inventory element deserved more attention than its single bullet in the alert received.

Instructions an employee can act on

The alert’s examples of detailed instructions covered penetration testing, security monitoring, access rights and reporting. On reporting, the stronger policies specified what to do, including whom to contact, if sensitive information was lost, stolen or misdirected. A policy either passes or fails that test. An employee who emails a client file to the wrong address needs a name and a step.

Tailoring cuts the other way too. A very narrowly scoped policy fails as surely as a vague one. A data protection policy that covers only the trading platform leaves the email system, the document repository and the vendor portals unaddressed. The staff listed narrow scope as its own failure mode.

The cost of tailoring

Tailored policies cost more to write and more to maintain. Each named system, role and interval becomes something that can go stale. When a firm replaces a CRM, the access policy has to change. When it reorganizes, the approvers change. A generic policy never needs updating because it never says anything specific. Its vagueness is both its appeal and its weakness.

Written for staff, read by examiners

A policy has two audiences. Staff need to know what to do. Examiners need to know what the firm committed to. Firms that write for the examiner alone produce documents full of principles and short on steps. Firms that write for staff alone produce runbooks with no statement of intent. The alert’s complaint about vague, general policies describes the first failure. It did not say how often each failure occurred.

One test is to hand a procedure to someone who has not worked on it and ask them to perform it. If they cannot tell which system, which form or which approver is meant, the procedure is not tailored. That test costs little. It also finds contradictions between documents before an examiner does.

A firm can manage this by separating layers. A short policy states principles and assigns ownership. Procedures name systems and steps. Standards or runbooks hold parameters such as review intervals. The alert did not prescribe this structure. It is one way to meet the staff’s concern without rewriting the whole program each quarter.

The adherence gap: policies the firm did not follow, or practices it never wrote down

The second issue was adherence. Firms did not appear to adhere to or enforce their policies, or the policies did not reflect actual practice [1]. The staff gave four examples. Each is a comparison between a document and a record.

Four examples, four comparisons

First, policies required annual customer protection reviews, but the reviews happened less frequently. Second, policies required ongoing reviews of whether supplemental security protocols were appropriate, but reviews were annual or did not happen. Third, policies gave contradictory instructions, such as remote customer access rules that seemed inconsistent with rules for investor fund transfers. Fourth, policies required all employees to take security awareness training, but firms did not appear to ensure completion or act on non-completion.

Each example could be found by asking for evidence. An examiner asks for the last three annual reviews and receives one. An examiner asks for training completion records and finds no follow-up on people who skipped it. The method is simple. The alert said the second sweep involved more validation and testing than earlier work, and testing finds this kind of gap.

Commitments a firm makes to itself

Several of these failures involved commitments the firm chose. As we read the alert, the annual interval in the first example came from the firm’s own policy, which is how the staff described it. A firm that writes a stricter commitment than it can keep creates a finding it would not otherwise have. A realistic interval that is met is easier to defend than an ambitious one that is missed.

That has a practical consequence. Before a policy is approved, someone should ask who performs each recurring task, what record it leaves and who checks that the record exists. If no one can answer, the commitment should change or the resourcing should. The alert’s list of stronger practices included senior management vetting and approving policies. Approval is the moment to ask that question.

Informal practice for funds transfers

The alert made a related point about money movement. The vast majority of broker-dealers and nearly two-thirds of advisers and funds, the staff reported, had authority from customers to transfer funds to third-party accounts. Some broker-dealers did not appear to put their processes into written supervisory procedures. They relied on informal practices to verify a customer’s identity before acting on a transfer request.

The staff reported that all advisers and funds examined had written policies on verifying the authenticity of a customer asking to transfer funds. The contrast is instructive. An informal practice may work well when the same experienced staff handle every request. It is hard to supervise, hard to train and hard to evidence when someone new is on the desk.

Third-party transfer authority is where a cyber incident becomes a financial loss. A spoofed email or a compromised client mailbox leads to a wire request. The control that stops it is a callback or other verification step. If that step lives only in habit, it can be skipped under pressure. Writing it down allows supervision and testing.

Contradictions between policies

The third example, contradictory instructions, points to a governance problem. The alert did not say how the contradiction arose. One possibility is that remote access rules and fund transfer rules were written by different groups at different times, with no one reading them together. The staff said the result made it unclear to employees whether certain activity was permissible. A firm can address this with a single owner for the policy set and a periodic cross-reading.

Monitoring as the missing control

Each adherence example describes a control that existed on paper and a check that did not. No one compared the review schedule with the reviews performed. No one compared the training roster with completions. The absent element in each case was compliance monitoring: a second line that tests whether the first line did what the policy says.

Advisers already had a hook for this. The fiscal 2027 priorities list an adviser’s annual compliance review among the items examinations analyze [5]. A cybersecurity section in that review can sample recurring tasks, compare them with records and report exceptions. Broker-dealers have their own supervisory review structures that can carry the same tests.

Monitoring also changes what an examiner finds. A gap that monitoring detected, recorded and fixed shows a working program. The same gap found first by the examiner shows that the program did not see it. The underlying lapse is identical. The finding is not.

Training as evidence

The training example is the easiest to fix and the easiest to miss. Learning platforms record completion. The gap the staff described was enforcement: firms did not appear to act on employees who did not finish. The stronger firms, the alert said, had mandatory training at onboarding and periodically after, with procedures to ensure completion. A completion report and a documented escalation close the gap.

The maintenance gap: patches, unsupported systems and test findings left open

The third part concerns system upkeep. The staff tied these issues expressly to Regulation S-P, describing them as Regulation S-P-related issues among firms that did not appear to adequately conduct system maintenance [1]. In our reading, that framing matters. It placed patching and remediation inside a privacy and safeguarding rule, as well as among general operational expectations.

Two named failures

The alert named two examples. The first it labeled stale risk assessments: firms using outdated operating systems that no longer received security patches. The second it labeled lack of remediation efforts: high-risk findings from penetration tests or vulnerability scans that did not appear to be fully remediated in a timely manner.

The label on the first is worth reading closely. The staff did not call it a patching failure. It called it a stale risk assessment. In our reading, the staff was saying that a current risk assessment would have flagged an unsupported operating system as a risk, and the firm would then have had to address it. The failure was in the assessment as much as in the technology.

Finding is not fixing

The remediation example shows the limit of testing. The staff reported that nearly all broker-dealers ran penetration tests and vulnerability scans on critical systems. Running a test produces a list. If high-risk items on that list stay open, the firm has documented its own exposure. An examiner who reads the report and then asks for the remediation record has an easy finding.

Remediation tracking is an operational discipline. Each finding needs an owner, a severity, a target date and a closure record. Findings that cannot be fixed need a documented risk acceptance or a compensating control. The alert did not set timelines. The absence of a timeline puts the burden on the firm to set one and meet it.

Patching with care

The alert did not ask firms to patch recklessly. Its list of stronger practices described patch management policies that included beta testing a patch with a small number of users and servers before firm-wide deployment. They also included analysis of the problem the patch fixed, the risk of applying it, and the method of applying it.

That balance matters for firms with trading or order-routing systems. A bad patch on a trading platform during market hours is its own incident. The answer the staff described is a documented process that weighs both risks, not a blanket rule. The record of that weighing is the evidence an examiner can review.

Where vendors hold the systems

Maintenance is harder when vendors run the systems. An adviser using a hosted portfolio platform cannot patch it. It can require the vendor to patch and to report. The alert noted that almost all firms either assessed vendor risk or required vendors to provide audit reports, security reviews or certification reports. Over half also updated vendor risk assessments at least annually, the staff reported.

The stronger practices included requiring third-party vendors to provide logs of their activity on the firm’s networks. That is a different control from a vendor questionnaire. It lets the firm see what a vendor’s staff actually did with their access. It extends the adherence logic to vendors: the question is what the record shows, not what the contract promises.

Measures a firm could keep

Maintenance lends itself to measurement. A firm can track the count of open high-risk findings by age, the share of critical patches installed within its own target, and the number of systems on unsupported software. The alert required none of these. Each one turns a maintenance commitment into a number that management can see and an examiner can request.

Measures need definitions to mean anything. A firm should state what counts as critical, when the clock starts and what counts as closed. A patch applied to a test server but not to production is not closed. A finding marked risk accepted without an approver is not closed either. Clear definitions keep the measures honest over time.

Closing the loop to the risk assessment

The staff’s label, stale risk assessments, implies a loop. Test results and patch status should feed the next risk assessment. The risk assessment should then change priorities for testing and patching. When the loop breaks, each activity runs on its own calendar and none informs the others.

A firm can check the loop by reading its latest risk assessment beside its latest test report. If the test found a high-risk issue in a system the assessment rated low, one of them is wrong. If the assessment does not mention unsupported systems the inventory lists, it is stale in exactly the sense the alert described.

Unsupported software as a business decision

Unsupported operating systems persist for reasons. A legacy application may run only on an old platform. Replacing it may be expensive or may depend on a vendor’s roadmap. The staff’s framing suggests what an examiner expects in that case. The firm should know the system is unsupported, assess the risk, isolate the system if possible, and record a decision and a plan.

The response gap: plans for incidents and customer notice that some firms lacked

The fourth part concerns what happens when controls fail. The alert reported that nearly all firms had plans for access incidents, and the vast majority had plans for denial of service incidents and unauthorized intrusions [1]. The weaker area was data breach and customer notification planning, and it differed by type of firm.

A difference between firm types

According to the staff, the vast majority of broker-dealers kept plans for data breach incidents, and a majority had plans for notifying customers of material events. Less than two-thirds of advisers and funds appeared to keep such plans. That difference is the clearest firm-type split in the alert.

The alert did not explain the difference. One possible reading is structural. An adviser or fund that outsources administration and shareholder records may have assumed its providers would plan for breaches on its behalf. We offer this as a hypothesis, not a finding from the alert.

Why customer notice planning was hard in 2017

In our reading, Regulation S-P in 2017 required safeguarding policies but set no customer notice duty. The SEC added one only in 2024, and its release said the amendments would help ensure customers receive notice in cases where state law did not require it [2]. A firm planning for notice in 2017 had to work out which state laws applied to its customers. In our reading, that complexity explains part of the gap.

A notice plan has several parts. It needs a decision process for whether an incident triggers notice, a way to identify affected customers, template content, a channel, and a record of what was sent. Each part depends on earlier work. Identifying affected customers depends on the data inventory from the first part. Without the inventory, notice planning is guesswork.

Roles and responsibilities

The alert reported that all broker-dealers and a large majority of advisers and funds kept cybersecurity organizational charts or described cybersecurity roles. Named roles are a necessary base for response. When an incident starts, someone must decide, someone must investigate, someone must communicate, and someone must document. A chart that names those roles before the incident saves hours during it.

Response planning and vendors

Response plans for firms that outsource data handling have to reach the vendor. If an administrator suffers a breach involving fund shareholders, the fund needs to learn of it quickly. In 2017 that depended on contract terms. The alert did not discuss vendor notice. As the hindsight section shows, the SEC later wrote a vendor notice requirement into Regulation S-P itself.

Account takeover as the common case

The alert’s response categories were access incidents, denial of service, intrusions and data breaches. For a retail brokerage or advisory firm, a real event can combine them. Someone takes over a customer’s email or account credentials and tries to move money. That event tests the transfer verification procedure from the second part and the response plan from this one at once.

A response plan for that case needs specific steps. Freeze or flag the account, verify with the customer through a known channel, check for other affected accounts, preserve evidence and decide on notice. Each step can be written down and practiced. The Regulation S-ID red flags program the alert cited is the natural home for the detection side.

Testing the plan

A response plan is a policy like any other, so the adherence logic applies. A plan that has never been exercised is a document with untested assumptions. Tabletop exercises produce records: who attended, what scenario ran, what gaps appeared, and what changed afterward. Those records are what an examiner can review. The alert did not require exercises, but its testing method makes them the natural evidence.

How the alert reached different firms

The alert covered three kinds of SEC registrant. Its findings and their consequences landed differently on each, and on firms outside its scope that shared the same infrastructure.

Broker-dealers

Broker-dealers came out ahead on several measures in the staff’s counts: risk assessments, penetration testing, breach plans and organizational charts. Their weak point in the alert was informal transfer verification. For a broker-dealer, the follow-up work centered on written supervisory procedures for money movement and on remediation records for test findings.

Investment advisers

Advisers showed weaker results on penetration testing and breach planning, according to the staff. Advisers range from single-office firms to large asset managers. A small adviser may have no security staff and rely entirely on vendors and an outside IT provider. For that firm, the alert’s practical meaning was vendor oversight and a written, simple, tested response plan.

Registered funds

Funds operate through service providers, such as advisers, administrators, transfer agents and custodians. In our reading, the alert’s findings for funds were in large part findings about the oversight of service providers’ controls. A fund chief compliance officer’s follow-up could include collecting and reviewing evidence from each provider.

Dual registrants and bank affiliates

Firms registered as both broker-dealer and adviser faced both sets of expectations. So did broker-dealer and advisory affiliates of banks. A bank holding company affiliate might share infrastructure with the bank, which was examined by banking agencies under different guidance. The same patching backlog could appear in a bank examination and an SEC examination. Aligning evidence across both reduced duplicated work.

Vendors serving the industry

Technology providers to advisers and broker-dealers were not examined in the sweep. Their clients were. A client that read the alert could ask its vendors for patch records, test results and incident procedures. In our reading, the alert pushed a share of the evidence burden onto vendors through their clients’ oversight programs.

Senior management and boards

The last element on the staff’s list was engaged senior management: policies vetted and approved by senior management. Approval creates accountability for the commitments a policy makes. It also creates a moment to test feasibility, as the second part described. Senior managers who approve a policy should also see the monitoring results that show whether it is followed.

For funds, the equivalent audience includes the board that oversees service providers. For a broker-dealer or adviser inside a bank holding company, it may include a risk committee at the parent. The alert did not address boards. Its logic suggests that whoever approves the policies should receive the evidence that they work.

Firms the alert did not cover

Banks, credit unions and health systems were outside the alert. We would still apply the method it described to them, because any examiner or auditor can compare a policy with a record. The four gaps in this article are not specific to securities firms. The rule names differ, but the evidence questions are the same.

What an evidence test leaves out

The alert’s method has limits, and firms should know them. A record shows that an activity happened. It does not show the activity was done well. A completed access review may have rubber-stamped every entitlement. A training completion may reflect someone clicking through slides. An evidence test rewards producing records, and a firm can produce records without producing security.

Examinations are also periodic. A sweep covering a fixed review period sees a slice of a firm’s history. Controls that work well during an examination year and lapse afterward will not appear in the findings. That argues for internal monitoring that runs all year, rather than a burst of preparation before an examination.

Finally, the method favors firms that can afford documentation. A large firm can staff a control testing team. A small adviser may have a chief compliance officer who also handles operations. For that firm, the right response is a smaller set of commitments it can evidence, not a large set it cannot. The alert’s concern was the gap, not the size of the program.

These limits do not weaken the core point. A firm’s records are still the first thing an examiner reads. The limits mean a good record is necessary but not sufficient. Firms should use the evidence test to find gaps and then ask whether the activity behind each record achieved its purpose.

Objections, in order

The alert said so itself. The Commission expressed no view on it, and it created no safe harbor. Yet its conclusion said OCIE would continue to examine for cybersecurity compliance procedures and controls, including testing their implementation. A deficiency letter citing a firm’s failure to follow its own policy rests on existing rules, not on the alert. The alert told firms where examiners would look. That changed the cost of ignoring the gap, even without changing the law.

2. Seventy-five firms cannot represent the industry

The sample was small compared with the number of registrants. The staff did not claim it was representative and gave no statistical design. We do not treat its proportions as industry rates. The argument here rests on the method and the categories of gap, which do not depend on the sample being representative. Any firm can test itself against them.

3. Some gap between policy and practice is unavoidable

That is true, and the alert did not claim otherwise. Its concern was gaps a firm did not know about or did not manage. A missed review that is detected, recorded and corrected is a functioning program. A missed review no one noticed is a finding. The difference is monitoring, and monitoring is within a firm’s control.

4. The alert read new expectations into a vague rule

Regulation S-P’s safeguards rule was short and general in 2017. One could argue that tying patching and remediation to it stretched the rule. In our reading, the safeguards rule already required reasonable design, and an unpatched, unsupported system is hard to call reasonably designed. Firms that disagreed had a forum in the examination process. They did not have a basis to ignore the risk.

5. Perfect patching is impossible

Firms cannot patch every system immediately, and some systems cannot be patched at all. The alert recognized this by describing testing before deployment. Its criticism was of high-risk findings left open and unsupported systems left unassessed. A documented decision with a compensating control is a different record from silence. The objection describes a constraint, not a defense.

6. Small firms cannot carry this documentation load

A single-office adviser cannot match a large broker-dealer’s control testing. The alert did not ask it to. Its stronger practices were examples, and its disclaimer said some factors may not apply to a given firm. The adherence logic scales down well. A short policy with few commitments, each met and recorded, passes the same test a large program does.

What a chief compliance officer could check

The alert’s method can be applied internally. Each check below compares a document with a record, as the staff did.

  • Pick three recurring commitments in the cybersecurity policies and request the last three records of each. Note any missing.
  • Compare the data and vendor inventory with the systems named in the access, patching and incident policies. List systems that appear in one and not the other.
  • Pull the open findings from the latest penetration test and vulnerability scan. Check each has an owner, a date and a status.
  • List systems running unsupported software. Check each has a recorded risk decision and a compensating control.
  • Read the written procedure for verifying third-party transfer requests. Walk one recent request through it and check the evidence.
  • Pull training completion records and the escalation record for anyone overdue.
  • Check that the incident response plan names a decision-maker for customer notice and has been exercised in the last year.
  • Ask the three largest vendors holding customer data for their notice commitments and their latest patching evidence.

These checks do not prove a program is effective. They show whether the program leaves the evidence an examiner would ask for. In our reading, that was the measure the August 2017 alert introduced.

With hindsight

Later SEC actions moved several of the alert’s concerns from staff observation into rule text, and dropped others. The pattern is easiest to see by following Regulation S-P, the rule the staff tied to its maintenance findings.

On May 16, 2024, the Commission adopted amendments to Regulation S-P. The final rule was published in the Federal Register on June 3, 2024 and took effect on August 2, 2024 [2]. It required broker-dealers, investment companies, registered investment advisers, funding portals and transfer agents to adopt written policies and procedures for incident response programs.

The response program had to address unauthorized access to or use of customer information. It had to include notice to affected individuals whose sensitive customer information was, or was reasonably likely to have been, accessed or used without authorization. The rule set that notice for as soon as practicable, and no later than 30 days after the firm becomes aware of the incident [2].

The amendments also reached vendors. Firms’ policies had to be reasonably designed to ensure service providers give notice of certain breaches. A provider must notify the firm as soon as possible, and no later than 72 hours after becoming aware of a breach affecting a customer information system it maintains. The release said the 72-hour figure was informed by the deadline in the Cyber Incident Reporting for Critical Infrastructure Act of 2022 [2].

The rule added requirements to keep written records documenting compliance. It extended safeguarding to transfer agents and broadened the information covered. It gave larger entities 18 months after Federal Register publication to comply and smaller entities 24 months [2]. In our reading, the response gap the 2017 alert found among advisers and funds became a rule requirement for the same types of firm.

A broader rulemaking went the other way. On March 9, 2022, the Commission had published proposed cybersecurity risk management rules for advisers, registered investment companies and business development companies. On April 5, 2023, it had published a proposed cybersecurity risk management rule for broker-dealers, clearing agencies, exchanges and other market entities. On June 17, 2025, the Commission formally withdrew both proposals, among others. It said it did not intend to issue final rules on them [3].

So the specific program requirements the 2022 and 2023 proposals would have imposed did not arrive. The general safeguards obligation, now with response and notice duties, did. The examination method carried the rest. On November 17, 2025, the Division of Examinations announced its fiscal 2026 priorities. It said it would examine for compliance with new rules, such as the 2024 amendments to Regulation S-P [4].

On October 23, 2026, the SEC announced its fiscal 2027 examination priorities [6]. The priorities document lists cybersecurity under information security and operational resiliency. Examinations will review policies on governance, data loss prevention, access controls, account management, and response and recovery for cyber incidents, including ransomware. They will also assess risks from registrants’ use of third-party service providers [5].

The same document says that, in line with the amended Regulation S-P, staff will evaluate response program policies, including customer notification procedures and oversight of service providers. Under Regulation S-ID, it says examiners will focus on red flags during customer account takeovers and fraudulent transfers [5]. The list overlaps closely with the six areas of the 2015 sweep and the transfer verification issue of 2017.

What held up is the method. The 2017 alert tested whether written policies were implemented and followed. The 2027 priorities describe examinations that focus on whether firms have developed, implemented and maintained reasonably designed policies. What changed is that notice and vendor duties now have dates and deadlines in rule text. The evidence question the alert asked in 2017 remains the one firms should expect.

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

  • SEC examinations
  • Regulation S-P
  • broker-dealers
  • investment advisers
  • patch management
  • incident response