The 2016 Barclays and Credit Suisse dark pool orders found protective features that did not run as marketed

The SEC's January 31, 2016 orders found that overrides, scoring and data feeds behind dark pool marketing differed from what subscribers were told. A description given to customers worked as a specification; a 2018 rule, Form ATS-N, made the main one public.

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 January 31, 2016 orders found that overrides, scoring and data feeds behind dark pool marketing differed from what subscribers were told. A description given to customers worked as a specification; a 2018 rule, Form ATS-N, made the main one public.

  1. The week The order also addressed a marketing chart.
  2. The description: what subscribers were told The Barclays order traced its descriptions through several channels.
  3. The behavior: what the systems did Liquidity Profiling was designed to sort subscribers by measured behavior. [2] The order said Barclays used metrics including 1-Second Take Alpha and Normalized Order Size. [2] The protection subscribers chose depended on those categories.
  4. The boundary: who could see subscriber data The finding turned on capability, not on proof of misuse.
  5. The change record: keeping claims current A 20-day advance filing means change management has to know, before release, that a change is material.
  6. How the findings reach different firms The direct lesson falls on firms that run ATSs.
  7. Objections, in order The orders contain both.
  8. Checks a platform owner could run Collect every customer-facing statement about what the platform does, across marketing, questionnaires and sales material.
  9. With hindsight The proposal pending that week became a rule.

The week

On January 31, 2016, the Securities and Exchange Commission announced settled cases against Barclays Capital Inc. and Credit Suisse Securities (USA) LLC over their operation of alternative trading systems. [1] The release said the cases found the firms violated federal securities laws while operating dark pools, and Credit Suisse’s Light Pool. [1]

The release said the New York Attorney General’s office was announcing parallel actions against both firms. [1] It said Barclays agreed to settle by admitting wrongdoing and paying $35 million penalties to the SEC and the New York Attorney General, for a total of $70 million. [1]

It said Credit Suisse agreed to pay a $30 million penalty to the SEC, a $30 million penalty to the New York Attorney General, and $24.3 million in disgorgement and prejudgment interest to the SEC. [1] It gave the total as $84.3 million. [1] Credit Suisse consented to the SEC’s orders without admitting or denying the findings. [1]

The Barclays order concerned LX, which the order said was, as of May 2014, the nation’s second largest NMS stock ATS. [2] The order said that from December 2011 through June 2014, Barclays made materially misleading statements and omissions about a feature called Liquidity Profiling and about its market data feeds. [2]

Barclays, the order said, designed Liquidity Profiling to sort subscribers’ order flow into categories from 0, the most aggressive, to 5, the least. [2] Subscribers could then choose not to trade with subscribers in certain categories. [2]

The order said Barclays marketed Liquidity Profiling as a surveillance framework that protected clients from predatory trading. [2] It said Barclays told subscribers it ran surveillance reports each week to check that the pool held no toxic order flow. [2]

The SEC found that Barclays did not police LX continuously for predatory trading through the visualization tools it depicted, as it had claimed, and did not generally run surveillance reports weekly during the period. [2] It found Barclays did not adequately disclose that it overrode Liquidity Profiling’s categorizations, at times moving subscribers from the most aggressive buckets to the least aggressive. [2]

The order said the subscribers moved included a Barclays market making desk. [2] As a result, the SEC found, subscribers that had chosen not to trade with aggressive flow still interacted with it. [2]

The order also addressed a marketing chart. It said Barclays removed the bubble for its largest subscriber with aggressive trading characteristics in October 2012, kept a footnote saying the chart showed the largest participants in LX, and in April 2014 amended the footnote to give the data source as Barclays data from 2014. [2] The order said the chart had been created from December 2011 data. [2]

On data feeds, the order said Barclays calculated the National Best Bid and Offer (NBBO) in LX from SIP data plus direct feeds from only three exchanges. [2] It said LX never subscribed to a direct NYSE feed during the period. [2] It said an April 9, 2014 marketing piece told subscribers Barclays utilized direct feeds from exchanges to deter latency arbitrage, without mentioning the SIP. [2]

The Barclays order also found violations of the Market Access Rule. [2] It said Barclays’ primary control, a system called NICON, sent email alerts as clients approached credit thresholds but could not reject orders that would exceed them. [2]

One Credit Suisse order concerned Crossfinder, which it said was, as of March 31, 2015, the second largest dark pool by dollar volume. [3] The SEC found that Crossfinder accepted and ranked sub-penny orders from at least April 2009 through April 2011, and executed them, in violation of Rule 612 of Regulation NMS. [3]

That order also found Credit Suisse transmitted confidential subscriber order information from Crossfinder to two internal applications outside the ATS. [3] One was its smart order router. The other was a server that sent indications of interest to exchanges’ outbound routers. [3] It found Credit Suisse did not fully disclose those transmissions to all subscribers. [3]

The order found Credit Suisse misrepresented that its alpha scoring of subscribers was objective, transparent and performed monthly. [3] It found Credit Suisse told clients its router did not preference Crossfinder when it systematically prioritized Crossfinder in certain stages of dark-only routing. [3] It also found Credit Suisse did not disclose a technology called Crosslink, through which its router traded certain client orders against orders from two high frequency trading firms. [3]

The second Credit Suisse order concerned Light Pool. [4] The SEC found Credit Suisse represented that a scoring formula, which the order calls the Alpha Formula, would be in use when Light Pool became operational, but did not begin applying it until June 2012. [4] It found the formula was applied per system ID, so a subscriber scored as opportunistic could keep trading through other IDs. [4]

The SEC had a rule proposal on the table that week. On December 28, 2015, it had published a proposal to require NMS Stock ATSs to file a new Form ATS-N, and to make certain Form ATS-N filings public on the Commission’s website. [5] Comments were due by February 26, 2016. [5]

The January 2016 orders read at first like cases about conflicts of interest in dark pools. [1] Read for their technology findings, they are about something more general: the distance between what a firm told customers its systems did and what the systems did. [2] [3] [4]

Our argument is that the orders treated a description given to customers as a specification the firm’s systems had to meet. Marketing pieces, questionnaire answers and pitchbook charts became evidence when the code, the configuration or the operating routine did something else. [2] [3] For any firm that sells a platform with protective features, the description is part of the system.

That framing matters to technology leaders because the evidence the SEC used was technical. It turned on override records, configuration settings, data feed subscriptions, order tags and internal data flows. [2] [3] [4] Those are artifacts that engineering teams create and keep. When a claim fails, they are what a regulator reads.

We take the gap in four parts, each a different place where description and behavior can come apart:

  1. The description: what subscribers were told, and where they were told it.
  2. The behavior: what the systems and the people running them did.
  3. The boundary: who and what could see subscriber data.
  4. The change record: how changes reached the regulator and the marketing material.

The parts do not overlap. The first is the claim, the second the conduct, the third the data perimeter, and the fourth the process for keeping claims current. After them we look at how the findings reach different firms, take the objections in order, and end with checks a platform owner could run.

One note on the findings. Barclays admitted the facts in its order and acknowledged its conduct violated the federal securities laws. [2] Credit Suisse consented to its orders without admitting or denying the findings. [3] [4] We describe both as the SEC’s findings.

The description: what subscribers were told

Several channels, one claim

The Barclays order traced its descriptions through several channels. It cited a two-page marketing piece on Liquidity Profiling, responses to subscriber questionnaires, proactive marketing statements and pitchbooks whose covers were updated monthly. [2] It also cited an emailed marketing piece sent on April 9, 2014. [2]

Each channel carried a version of the same claim: that Barclays watched the pool and protected clients. [2] The order treated each as a statement the firm had to stand behind. [2] The legal hook was Section 17(a)(2) of the Securities Act, which the release said both firms violated. [1]

For a platform firm, the point about channels is practical. A product claim does not live only in the official disclosure. It lives in sales decks, questionnaire answers and emails, each written by different people at different times. Each is a statement about the system.

Why the protective claims existed

The Barclays order said Barclays developed Liquidity Profiling in December 2011 to differentiate LX from other dark pool operators. [2] The protective feature was a selling point by design. That gives it a different status from an internal tool. It was built to be described.

A feature built to be described carries a heavier evidence burden. Its marketing is part of its purpose. The team that owns it has to be able to show, at any time, that it works as the sales material says. In our reading, the orders treat that as the operator’s responsibility, not the sales team’s alone.

Specific claims create specific duties

The claims that caused trouble were specific. A promise of weekly surveillance reports is a testable statement. [2] So is a promise of continuous policing. [2] So is monthly alpha scoring. [3] So is a router that did not preference the firm’s own pool. [3]

Specific claims are what customers buy on. They are also what a regulator can check against logs. A firm that promises a weekly report needs a report for every week. The order said Barclays did not generally run weekly surveillance reports during the period. [2]

One claim, one owner

Each of the claims in the orders had a natural owner inside the firm: the team that ran the surveillance, the team that maintained the scoring, the team that configured the feeds. The material that carried the claims was something else: marketing pieces, pitchbooks, questionnaire answers and statements to clients. [2] [3] In our reading, that separation between the claim and the system is where drift begins.

A claim with no owner in the system team drifts. A sales deck can repeat a statement long after the process behind it has changed. A simple register that links each claim to the team that can evidence it, and requires that team’s sign-off before reuse, closes that drift.

The confidential filing

Regulation ATS required a filing too. The Barclays order described Form ATS as a confidential document an ATS uses to notify the Commission of its operations. [2] The Crossfinder order used the same description. [3]

That made the regulatory description invisible to subscribers. They relied on what the firm told them directly. The filing and the marketing could diverge, and subscribers would not see the difference. The Crossfinder order found that Credit Suisse’s Form ATS did not disclose that Crossfinder transmitted order information to its router. [3]

Words that need a number behind them

The Crossfinder order said Credit Suisse represented to clients that it analyzed routing venues on a fixed schedule or continuously, using words for monthly, regular, routine or continuous review. [3] The order said that before January 2013 the firm analyzed venues on an ad hoc basis. [3]

Each of those words implies a schedule. A schedule can be checked against a calendar of completed reviews. A firm that uses such words should be able to produce the calendar. Where it cannot, the safer word describes what actually happens.

Questionnaire answers as disclosures

The Barclays order described an answer to a subscriber’s question about data feeds: Barclays said it used a combination of SIP data for regional venues and direct feeds for major exchanges. [2] The order said Barclays never subscribed to a direct NYSE feed during the period. [2]

A questionnaire answer is written for one customer. The order treated it as a representation. For a platform owner, every due diligence questionnaire is a disclosure document. It needs the same review as public material.

The behavior: what the systems did

Overrides as a feature

Liquidity Profiling was designed to sort subscribers by measured behavior. [2] The order said Barclays used metrics including 1-Second Take Alpha and Normalized Order Size. [2] The protection subscribers chose depended on those categories.

Overrides changed the categories by hand. The order said the majority of overrides moved subscribers to more aggressive buckets, but some moved subscribers from the most aggressive to the least aggressive. [2] Those included Barclays’ own market making desk, which the tool at times rated 0 or 2 and Barclays reclassified as 4. [2]

In architecture terms, a manual override is a second decision path. The automated path did what was described. The manual path changed the outcome. Subscribers who opted out of aggressive flow were relying on the automated path. [2]

The order noted a footnote in some client reports in which Barclays reserved the right, as operator of LX, to override any participant’s profile. [2] It found that neither those reports nor other materials defined what an override meant. [2] Nor, it found, did they disclose overrides that let aggressive subscribers trade against those who had opted out. [2]

Scoring that was not what it seemed

The Crossfinder order found alpha scoring included significant subjective elements, was not transparent, and did not categorize all subscribers monthly, from December 2010 through March 2014. [3]

The Light Pool order found its scoring formula, the Alpha Formula, was not applied until June 2012, though Light Pool was fully operational for trading by at least June 2011. [4] It found the formula was applied per system ID rather than to a subscriber’s trading as a whole. [4]

The per-ID finding is an architecture detail with a policy effect. A subscriber is a business. A system ID is a connection. A control keyed to the connection can be avoided by using another connection. The order said subscribers scored as opportunistic could keep trading through other system IDs. [4]

A list that set the routing

The router finding shows how a simple data structure can carry a policy. The Crossfinder order said certain dark-only routing default settings used a numbered list of available venues, with Crossfinder in the first position. [3] It said that resulted in Crossfinder being prioritized during three stages of the dark-only routing process. [3]

The order said the routing methods produced a statistical bias, sending more client orders to Crossfinder than would have gone without its top position. [3] It said Credit Suisse told clients the router did not preference venues based on criteria other than execution quality. [3]

In architecture terms, the venue list was configuration, and configuration decided outcomes. A review of the routing code alone might not reveal the bias. A review of the default list would. The order said Credit Suisse recognized the bias for one stage as early as October 2013, and stopped using the list for the two ping stages in October 2014. [3]

Scoring as a promise

Both Credit Suisse orders concerned scoring: alpha scoring in Crossfinder and the Alpha Formula in Light Pool. [3] [4] In each, the firm described a measurement that would sort subscribers, and the SEC found the measurement was not applied as described. [3] [4]

A score that customers rely on to avoid certain counterparties is a control. Its method, its schedule and its exceptions are part of the control. The Crossfinder order’s finding that alpha scoring included significant subjective elements matters here. [3] A subjective step is not wrong in itself. Undisclosed, it changes what customers think they are buying.

Prices the rules prohibited

The Crossfinder order found Crossfinder accepted and ranked orders priced in sub-penny increments, contrary to Rule 612 of Regulation NMS. [3] It said the orders came from three order tags created for Crossfinder. [3]

The firm’s own compliance manual, the order said, stated it would accept orders that imply a sub-penny price but would round them to an allowable increment. [3] The order quoted the Crossfinder FIX specification, which showed an example producing an adjusted limit price of 100.011. [3]

That is the gap in its plainest form. The policy said round. The specification, given to subscribers, showed an unrounded price. The system accepted it. [3] The order said that between October 2009 and April 2011, approximately 23 percent of Crossfinder executions of orders using the three tags were at impermissible sub-penny prices. [3]

Data feeds and the reference price

The Barclays order said Barclays used the NBBO to price pegged orders in LX and as a basis for certain compliance decisions. [2] The feeds that built the NBBO were therefore part of the pool’s core behavior.

The order said Barclays used SIP data plus direct feeds from BATS, ARCA and NASDAQ, and no direct NYSE feed. [2] Its marketing said it used direct feeds from exchanges to deter latency arbitrage. [2] The statement described a design intent. The configuration was a mix.

The order also recorded the fix. It said that from October 2014, LX used a third-party vendor that calculates the NBBO from direct data feeds from all protected venues under Regulation NMS. [2]

The internal desk

The override finding involved the operator’s own trading. The order said a Barclays market making desk was at times rated 0 or 2 by Liquidity Profiling in 2012 and 2013, and Barclays reclassified it as 4. [2] It said that let the desk trade with subscribers that had opted out of the categories the tool had assigned it. [2]

That is where a manual path meets a conflict of interest. A control that can be overridden by the operator, for the operator’s own desk, needs an independent reviewer and a disclosure. The order found neither was adequate. [2]

Controls that alert and controls that block

The Market Access Rule findings show a different gap. The order said NICON sent emails to Barclays’ service desk when a client reached percentages of its credit threshold, such as 33, 50, 66 and 80 percent. [2] It said NICON could not reject orders that would exceed the threshold, and no other control did. [2]

An alert depends on a person reading it and acting in time. A block does not. The order found Barclays lacked controls reasonably designed to prevent the entry of orders that exceeded pre-set credit and capital thresholds. [2] In our reading, the requirement’s word “prevent” is what the alert-only control could not meet.

The order added that sales traders could enter orders under a placeholder acronym not tied to any client, so some orders were not counted against client thresholds. [2] A control can only measure what is tagged to it.

The boundary: who could see subscriber data

Access inside the firm

Regulation ATS required safeguards to protect subscribers’ confidential trading information, including limiting access to those employees who operate the system or are responsible for its compliance. [2] The order quoted that requirement from Rule 301(b)(10). [2]

The Barclays order found employees in two user groups could view LX orders and executions the day after they occurred, if they knew the relevant computer language and could work through Barclays’ computer systems. [2] It found Barclays did not limit access as the rule required. [2] It said Barclays took steps to limit access after a December 2013 internal audit. [2]

The finding turned on capability, not on proof of misuse. Access was available to people who should not have had it. In our reading, that is the standard a data perimeter is held to: who can see, as well as who did.

Day-after access is still access

The access the Barclays order described was to orders and executions the day after they occurred. [2] A firm might argue that historical data is less sensitive than live orders. The order did not accept that as a defense. It found the access itself fell outside the rule’s limit. [2]

For data owners, retention and access controls apply to history as well as to live data. Archives and reporting databases are within the perimeter.

Data flowing out of the pool

The Crossfinder order described two flows of subscriber data to systems outside the ATS. [3] One sent Crossfinder order information, aggregated by price level, to the firm’s smart order router. [3] The other sent information to a server that transmitted indications of interest to the outbound routers of two exchanges. [3]

The order said that when the Commission adopted Regulation ATS, it highlighted the importance of separating ATS functions from other broker-dealer functions. [3] It treated the transmission of order data outside the ATS as evidence of a violation of Rule 301(b)(10). [3]

For an architect, these are integrations. Each crossed a boundary the rule drew around subscriber data. A data-flow map that marked the ATS perimeter would have shown both crossings.

De-identified is not the same as confidential

The Crossfinder order said the order data sent to the router did not include subscriber identifying information. [3] It did include symbol, side, size and price. [3] The order still treated the transmission as evidence of a Rule 301(b)(10) violation. [3]

For data architects, this finding deserves attention. Removing names did not make trading information non-confidential under the rule. Order details are valuable in themselves to a router that decides where to send other orders. The perimeter protects the information as well as the identity.

Settings only the operator could change

The order said the router feed was one of several settings in each subscriber’s configuration, and that the setting could only be changed by Credit Suisse personnel. [3] It said a small percentage of subscribers were told, either because they asked or because they learned of the setting. [3]

A configuration that only the operator can change, and that customers are not told about, works as a hidden term. The order also said the router sat within a different business unit and its supervisor had no responsibilities over Crossfinder. [3] Ownership of the two systems did not meet. The data crossed between two units whose supervision did not overlap, by the order’s account. [3] A named owner for each data flow closes that kind of gap.

Defaults by customer type

The Crossfinder order found that indications of interest were sent by default for sell side subscribers and not for buy side subscribers. [3] It found Credit Suisse did not inform certain subscribers they could opt out of the indications of interest and the router feed. [3]

Defaults are design decisions with disclosure consequences. A default that differs by customer type, and that customers do not know they can change, is a term of service set in code.

Access limited unfairly

The Crossfinder order also found violations of the fair access rule, Rule 301(b)(5). [3] For securities where Crossfinder met the volume test the order described, it found Credit Suisse unreasonably limited access to several functionalities in an unfair or discriminatory manner. [3]

It said the firm did not tell certain subscribers they could opt out of the indications of interest or the router feed. [3] It said only the two exchanges’ outbound routers received indications of interest from the server. [3] It said the firm did not disclose its grants, denials and limitations of access in its Forms ATS-R. [3]

A side door for two firms

The Crossfinder order described Crosslink, which it said operated from May 2009 to December 2012. [3] It said the router exchanged messages with two high frequency trading firms that told them a particular order for a particular security, volume and price was available. [3] The order found Credit Suisse did not disclose that its router routed child orders for execution against those firms’ orders. [3]

The change record: keeping claims current

The 20-day rule

Regulation ATS, as the Barclays order described Rule 301(b)(2), required an ATS to file an amendment on Form ATS “at least 20 days before it implemented a material change.” [2] The Barclays order found Barclays instituted overrides throughout the period without amending its Form ATS. [2] The Crossfinder order found a similar failure for the indications of interest functionality. [3]

A 20-day advance filing means change management has to know, before release, that a change is material. That is a classification step inside the release process. If engineers and product owners do not flag a change, the filing never happens.

Marketing that did not change with the system

The pitchbook chart is the clearest example of stale material. The order said that from December 2011 to September 2012 Barclays updated the month and year on the pitchbook cover each month but kept the bubble chart built from December 2011 data. [2] It said Barclays removed one subscriber’s bubble in October 2012 and, in April 2014, amended the footnote to say the data was from 2014. [2]

That is a document control failure as well as a disclosure failure. Marketing material that shows system data needs a source, a date and an owner. When the data is refreshed, the chart is refreshed. When it is not, the label says so.

The chart as a data product

The pitchbook chart was a visualization of system data. [2] It needed the same controls as any data product: a source, a refresh schedule and a record of edits. The order said one subscriber’s bubble was removed while the footnote still described the chart as covering the largest participants. [2] An edit log would have shown the change and who made it.

Late filing for a new venue

The Light Pool order found Credit Suisse did not timely update its Form ATS to reflect changes to how the Alpha Formula was applied, first implemented in the fall of 2012. [4] It also found the Forms ATS did not disclose that a subscriber could keep direct access through other system IDs. [4] Taken with the other findings, the orders show the same pattern across three venues. The systems changed faster than the descriptions of them.

Who decides what is material

The 20-day filing rule turns on materiality. [2] In a release process, someone has to make that call before deployment. If the decision sits with engineers, they need criteria. If it sits with compliance, compliance needs to see changes early enough to judge them.

The orders do not describe how either firm made materiality decisions. They show the result: changes in operation that the Form ATS did not reflect. [2] [3] [4] A release checklist with a materiality question, answered by a named person, is the minimum control. The answer and its reasoning should be kept with the release record, so a later review can see why a change was or was not filed.

Remediation as process

The Barclays order listed remedial steps. It said Liquidity Profiling reviews of direct subscribers were now monthly, and overrides could only move subscribers to more aggressive categories. [2] It said Barclays had filed an amended Form ATS describing the override process. [2]

The undertakings went further. The order required a third-party consultant to review how Barclays creates, approves and disseminates marketing material for LX, including tracking to whom it goes. [2] It also covered reporting material changes on Form ATS and the safeguards for confidential trading information. [2] In our reading, the SEC treated marketing approval as a control process, reviewable like any other.

How the findings reach different firms

A broker-dealer that operates a trading platform

The direct lesson falls on firms that run ATSs. Their marketing, questionnaire answers and regulatory filings describe the same system and need a single owner. Their overrides need to be logged and disclosed. Their data flows need a perimeter. The orders show each failure in a specific form. [2] [3] [4]

A broker-dealer or wealth manager routing client orders

Firms that route client orders to dark pools rely on the venues’ descriptions. The orders show those descriptions could differ from behavior. [2] [3] A routing firm can ask venues for evidence behind specific claims: surveillance reports, scoring methods, data feed configurations. It can also compare execution outcomes against what the venue says it does.

A bank with a broker-dealer affiliate

A bank holding company whose broker-dealer runs a platform carries the reputational and legal exposure the orders describe. Its board and risk functions see the broker-dealer through reports. The orders suggest those reports should include whether product claims match system behavior.

An institutional investor, including public funds

The Barclays order said LX subscribers included institutional investors. [2] The Crossfinder order described buy side firms as generally including hedge funds, mutual funds and pension funds. [3] Public pension funds that trade through brokers may therefore have orders reaching such venues. Their oversight of brokers can include questions about where orders go and what protections the venues claim.

Internal audit as the first detector

The Barclays order said the firm took steps to limit access to confidential trading information in response to a December 2013 internal audit. [2] It said further process changes followed the internal audit and regulatory investigations. [2]

For any platform firm, that is a useful reminder. Internal audit can test claims against systems before a regulator does. An audit plan that samples customer-facing claims each year and asks for the evidence behind them is a direct application of these orders.

A firm that buys execution technology

Firms that buy execution algorithms or routers from brokers rely on descriptions of routing logic. The Crossfinder finding about a default venue list shows what to ask for: the defaults, the order of venues, and when they changed. [3] A buyer that receives only a narrative description cannot test it.

A fintech or platform company outside securities

The pattern reaches beyond securities. A fintech that markets fraud screening, a payments firm that markets monitoring, or a health technology vendor that markets access controls makes testable claims about its systems. The orders show how a regulator reads such claims against logs and configurations. Securities law supplied the hook here. Other regimes have their own.

Objections, in order

1. These were conflicts of interest, not technology failures

The orders contain both. In our reading, the router preference and the market making desk override are conflicts. [2] [3] The sub-penny tags, the alerting-only credit control and the data feed mix are technology facts. [2] [3] In our reading, the conflicts went unseen because the systems allowed behavior subscribers could not observe.

2. Subscribers are sophisticated and should test venues themselves

Subscribers could measure outcomes. They could not see overrides, internal data flows or confidential Form ATS filings. [2] [3] Testing from outside does not reveal a manual reclassification inside.

3. Marketing is not a disclosure document

The orders treated marketing pieces, questionnaire answers and pitchbooks as statements under Section 17(a)(2). [2] [3] In these orders, the distinction a firm might draw between marketing and disclosure did not hold. A firm that wants its marketing to carry less weight has one option: make it say less, and make what it says true.

4. Overrides are necessary

They may be. The Barclays remedy kept overrides but limited their direction: they could only move subscribers to more aggressive categories. [2] The issue was undisclosed overrides that weakened a protection subscribers chose.

5. The penalties were the point, not the controls

The release put the totals at $70 million for Barclays and $84.3 million for Credit Suisse. [1] The undertakings were about process: marketing approval, market access controls, Form ATS reporting and data safeguards. [2] Those are the parts that change how a firm operates.

6. The 2018 rule solved the problem

The 2018 rule made Form ATS-N public for NMS Stock ATSs and gave the Commission a review process. [6] It did not cover what a firm says in a sales meeting or a questionnaire answer. Those statements still have to match the system and the filing. A public filing makes divergence easier to spot. It does not prevent it.

7. Only the largest venues draw this scrutiny

LX and Crossfinder were large, by the orders’ accounts. [2] [3] Light Pool was not. The Light Pool order said that as of June 30, 2014, Light Pool was approximately the 21st largest equities ATS out of 42 by dollar volume. [4] The findings against it were of the same kind. Size did not decide whether a gap between claim and behavior mattered.

Checks a platform owner could run

Inventory the claims

Collect every customer-facing statement about what the platform does, across marketing, questionnaires and sales material. Assign each to a system owner.

Test each claim against evidence

For each specific claim, identify the log, report or configuration that proves it. If a claim says weekly, produce every week’s report. [2]

Log and review overrides

Record every manual change to an automated control, with who, when and why. Review overrides against what customers were told. [2]

Key controls to the customer, not the connection

Check whether controls apply per customer or per technical ID. The Light Pool finding turned on that difference. [4]

Map data flows across the perimeter

Draw the boundary around customer data and list every integration that crosses it. Confirm each crossing is disclosed. [3]

Prefer blocking controls where the rule says prevent

Where a requirement uses the word prevent, check that the control blocks rather than alerts. [2]

Tie releases to disclosure review

Add a step in release management that asks whether a change is material to customers or regulators, before it ships. [2] [3]

Let customers see their settings

Give each customer a view of the configuration settings that affect how its orders or data are handled, and how to change them. The Crossfinder order found a setting only the operator could change, which it said a small percentage of subscribers were told about. [3]

Review default lists

Review default lists and priority orders in routing or decision logic as policy, with an owner and a stated rationale. [3]

Ask internal audit to sample claims

Have internal audit test a sample of customer-facing claims against evidence each year. [2]

Check reference data inputs

List the data sources behind any reference price or score the platform uses, and compare them with what customers are told. [2]

Prepare for the questionnaire

Keep a reviewed library of answers to standard due diligence questions, each tied to evidence and an owner, and refresh it when systems change. [2]

Date the marketing

Give every chart and statistic in sales material a source and a date, and an owner who refreshes it. [2]

With hindsight

The proposal pending that week became a rule. The SEC adopted amendments to Regulation ATS for NMS Stock ATSs, published in the Federal Register on August 7, 2018 and effective October 9, 2018. [6]

The final rule adopted Form ATS-N, which requires NMS Stock ATSs to disclose information about their manner of operations, the broker-dealer operator, and the ATS-related activities of the operator and its affiliates. [6] It required public posting of certain Form ATS-N filings through EDGAR. [6]

It also provided a process for the Commission to review Form ATS-N filings and, after notice and opportunity for hearing, declare a Form ATS-N ineffective. [6] It required all ATSs subject to the regulation to put in writing their safeguards and procedures to protect subscribers’ confidential trading information. [6]

The adopting release cited the January 2016 orders. A footnote said that since the proposal, the Commission had entered additional settlements regarding NMS Stock ATSs, and listed the Barclays order and both Credit Suisse orders. [6] It also listed a December 16, 2016 settlement with Deutsche Bank Securities Inc. [6]

The proposal had called for posting Form ATS-N filings publicly. [5] The January 2016 orders had described Form ATS as a confidential document. [2] [3] The final rule kept public posting for NMS Stock ATSs. [6]

The release explained the disclosure gap it addressed. It said NMS Stock ATSs, unlike national securities exchanges, were not required to publicly disclose their operations and fees. [6] It said that difference may create a competitive imbalance between two functionally similar trading centers. [6]

It set a schedule. An NMS Stock ATS operating under a Form ATS as of January 7, 2019 was required to file a Form ATS-N no earlier than January 7, 2019 and no later than February 8, 2019. [6]

The SEC later tried to extend parts of the framework. On December 31, 2020, it published proposals on Regulation ATS for ATSs that trade U.S. government securities and other matters. [7] On March 18, 2022, it re-proposed parts of them, together with amendments to Exchange Act Rule 3b-16, which defines terms in the statutory definition of exchange. [7] It reopened the comment period in May 2022 and again in May 2023. [7]

On June 17, 2025, the SEC withdrew the March 2022 proposal, along with other proposals. [7] It said it does not intend to issue final rules on them, and would issue a new proposal if it decides to pursue the area. [7]

The NMS stock framework remained. As checked on November 16, 2026, the eCFR text of 17 CFR 242.304(b), up to date as of November 12, 2026, says every Form ATS-N filed is a “report” under the Exchange Act. [8] It says the Commission will make public, via posting on its website, each effective initial Form ATS-N, as amended, and Form ATS-N amendments, on the terms the rule sets. [8]

What held up is the idea in the 2016 orders that an operator’s description of its system is a statement it must stand behind. The 2018 rule made the main description public and reviewable for NMS stock venues. [6] [8] What did not happen is the 2022 proposal to extend Regulation ATS requirements to more trading systems, including those trading government securities. The SEC withdrew it in 2025. [7]

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

  • dark pools
  • Regulation ATS
  • Form ATS-N
  • broker-dealers
  • market access