Seven regulators compared

The seven agencies' records differ in kind. Of 3,273 records, the SEC holds 1,844, SEBI (India) 656, the CFTC 334, the AMF 209, ASIC 113, the OSC 69 and the FCA 48; 122 of SEBI's are dismissals. Medians, unconverted: US$173,436 (SEC), US$1.5 million (CFTC), A$65,000, C$267,500, £279,922, €355,000 and INR 4.5 million (SEBI). Gaps follow collection as much as regulators.

Published 2026-10-05 · updated 2026-10-07 · 12 min read

Put the seven agencies’ records side by side and they look different in almost every respect: what they cover, how the matters ended, how often money is stated, and how large that money is in its own currency. Much of that difference is real enforcement behaviour. Some of it is simply how this library collected each agency’s records. This post separates what the data can say from what it cannot.

The finding first

The library holds 3,273 records from the seven. The SEC accounts for 1,844 (56%), the Securities and Exchange Board of India (SEBI) 656, the CFTC 334, the French Autorité des marchés financiers (AMF) 209, ASIC 113, the Ontario Securities Commission (OSC) 69 and the UK Financial Conduct Authority (FCA) 48. No other agency appears. The AMF’s records are decisions of its Commission des sanctions, the administrative sanctions committee, added on 5 October 2026. SEBI’s are orders of India’s securities regulator, added on 8 October 2026 (489 records) and 9 October (167 more) and described in their own section below.

• Money. A stored penalty figure exists for 90% of AMF records, 73% of CFTC, 70% of SEBI, 63% of FCA, 55% of SEC, 46% of OSC and 31% of ASIC. Counting any money field (penalty, disgorgement or interest), the SEC figure rises to 69% and the CFTC to 76%; for the AMF and SEBI it stays at 90% and 70%, because the library stores only one figure for them.
• Outcome. CFTC, ASIC, OSC, AMF and SEBI records are mostly judgments (61%, 85%, 54%, 80% and 71%); the FCA’s are mostly settlements (58%). The SEC’s largest group is settled (56%), with 8% still recorded as filed. SEBI has the largest share of dismissals, 122 of its 656 (19%), against 9% at the AMF, 4% at ASIC, 3% at the OSC and 1% at the SEC.
• Bans. 36% of OSC records record a ban, 31% of SEC, 30% of CFTC, 25% of SEBI (164 records), 9% of ASIC (10 records), 6% of FCA (3 records) and 2% of AMF (4 records).
• Criminal case alongside. 81% of ASIC records, 32% of CFTC and of SEC, 19% of OSC, 17% of FCA, 3% of AMF and 0.2% of SEBI (one record). For SEBI and the AMF, a false flag means the order does not report a criminal case, not that none exists; the criminal-parallel post explains.
• Typical penalty, each in its own currency. Median US$173,436 for the SEC (1,005 records) and US$1,500,000 for the CFTC (244). Median A$65,000 for ASIC (35), C$267,500 for the OSC (32), £279,922 for the FCA (30), €355,000 for the AMF (189) and INR 4,500,000 (Rs 45 lakh) for SEBI (462). These numbers are not comparable across currencies, and the next section says why.

How it was measured

A script, scripts/analysis/cross-border.mjs, reads every case file and groups by the agency field. Every figure in this post comes from its output.

Currencies. The SEC and CFTC amounts are the penaltyUsd field. For ASIC, the OSC, the FCA, the AMF and SEBI the amount is the penaltyNative field, whose currency occurs as AUD on all 35 ASIC records that have one, CAD on all 32 OSC records, GBP on all 30 FCA records, EUR on all 189 AMF records and INR on all 462 SEBI records. No record of the SEC or CFTC has a native amount, and no other currency appears. Nothing was converted. A conversion needs an exchange rate, and no single rate fits 2010 to 2026: sterling, the euro, the rupee and the Australian and Canadian dollars each moved by tens of percent against the US dollar across the period, and a rate fixed to one date would misstate the rest. The records span different years and different legal regimes, so even a correct conversion would not make a penalty under one regime comparable with another. The library’s own decision on this is the same: non-US penalties are recorded as published, and left out of every USD total. So this post compares shares and shapes, and within each currency the distribution of amounts. Indian amounts are given in rupees; one lakh is 100,000 and one crore is 10 million.

As of 7 October 2026. The AMF’s 209 records were added on 5 October, taking the library from 2,428 to 2,637 records. Since then 21 records that were not individual enforcement matters (16 SEC subpoena-enforcement applications, three CFTC umbrella announcements, a CFTC notice and a dismissed OSC motion) were removed, which lowers the SEC by 16 and the CFTC by 4 records, and SEBI’s 489 were added, making 3,106; on 9 October 167 more SEBI records were added, making 3,273. The medians for the SEC and CFTC did not move. A second reader re-read 60 AMF records on 7 October and the AMF figures reflect the corrections that followed. SEBI had a sampled second read of its first 489 records on 8 October (below); the 167 added on 9 October have no second read. Everything below was recomputed on the 3,273.

As of 4 October 2026 (evening). The figures were recomputed after the library’s status and flag corrections that day. Of the 733 records that had been recorded as filed, 543 were researched and moved to settled, judgment or another outcome, so the SEC’s “filed” count fell from 589 to 157 (147 now) and the CFTC’s from 107 to 30; the settled and judgment shares, the share of records with money and the ban shares moved with them. The criminalParallel flag was corrected on a further batch of records, and 805 of the then 2,428 records carried it (811 of 3,273 now). Five ASIC records that had held an Australian-dollar amount in a US-dollar field were fixed, and ASIC’s penalty count is now 35 where an earlier version of this post counted 26. On eleven further records the duplicated money on twin records of one matter was removed, which lowered the SEC and CFTC penalty counts slightly. The CFTC’s largest record, the five-bank benchmark order, is now stored at US$1.475 billion and as settled, and two further CFTC benchmark orders (Lloyds and R.P. Martin) moved from judgment to settled. Outcomes in some records cover only some of the defendants.

A penalty here is whatever monetary sanction the record stores in that field. For the OSC it includes the voluntary payments the Royal Bank of Canada and Toronto-Dominion settlements describe, which are payments to the regulator rather than penalties imposed after a finding; for ASIC it includes infringement-notice penalties; for the AMF it is the fine (sanction pécuniaire) the committee imposes or the settlement amount it homologates, and a “judgment” at the AMF means a sanctions decision rather than a court judgment. For SEBI it is the monetary penalty an order imposes or, for a settlement, the amount paid under settlement terms; SEBI orders also direct disgorgement of gains with interest, which the records describe in text but the library has no rupee field to store. Where a record involves several respondents the stored amount is the sum across them, and for SEBI some of those sums were made by the reading agents adding up the table of noticees in an order.

“Share with money” counts records in which a penalty is stored; I also give the wider count that adds disgorgement and interest, which the ASIC, OSC, FCA, AMF and SEBI records do not store (their native field holds the one figure). A record with no money may be a matter still pending, a criminal case, a dismissal, a settlement whose release gave no figure, or a record where the document was not clear enough to extract one.

What each agency’s records cover

The records are not a census of any regulator. The collection differs by source:

Four consequences matter for any comparison. First, each source is filtered by a classifier that keeps a matter only if it matches a manipulation-related technique, so the library is a selection of each agency’s output, not all of it. The FCA publishes far more final notices about consumer credit and fees than about market abuse, and the Tribunal’s docket is mostly unregistered trading, fraud and disclosure failures; those are read and skipped. Second, the Canadian coverage is Ontario only. British Columbia, Alberta and Quebec are not collected, and CIRO, which handles order-book conduct on Canadian marketplaces, cannot be reached by an automated client. FINRA is excluded for a different reason: its robots.txt disallows the disciplinary database. Third, the ASIC and FCA records come from media releases and final notices, not from court judgments, so a court outcome appears only if a release reports it. Fourth, the AMF records are the committee’s own decisions, which are administrative: an appeal to the Paris Court of Appeal or a criminal prosecution of the same conduct is not part of the record unless the decision mentions it. Older decisions are anonymised by the AMF (names appear as letters such as X or A), and the records keep those names as published; 154 of the 209 AMF records have a defendant recorded that way by a simple pattern check. This AMF is the French regulator, not the Québec body of the same initials, which is not collected. The SEBI records are also administrative, and are described in the next but one section. The sources page and the project’s decision log state each of these in more detail.

The SEC’s litigation releases begin in 2015 and its administrative orders in 2013, so the SEC’s early years are thinner, and 149 SEC records date from 2013 or 2014 against 1,685 from 2015 on. The CFTC, OSC and FCA start in 2013, ASIC in 2014, the AMF in 2004 (107 of its 209 records are from before 2014 and 101 from 2014 to 2025) and SEBI in January 2022. Excluding the part-year 2026, ASIC has 5 to 13 records a year, the OSC up to 10 (none in 2024), the FCA 1 to 9, the AMF 3 to 17 and SEBI 68 to 312, so none of this should be read as a trend.

What each agency’s records show

Technique mix. The SEC records are dominated by insider trading (533 records, 29%) and Ponzi schemes (337, 18%), followed by unregistered distributions (157, 9%) and pump and dump (111, 6%). The CFTC’s are different: Ponzi schemes (109, 33%), spoofing (72, 22%) and price manipulation (55, 16%) lead, so 44% of CFTC records fall in the order-book family against 5% for the SEC. ASIC is led by insider trading (55, 49%) and price manipulation (36, 32%). The OSC is split between insider trading (18, 26%) and Ponzi schemes (17, 25%). The AMF is the most concentrated on insider dealing: 120 of its 209 records (57%), then misleading issuer disclosure (38, 18%) and price manipulation (24, 11%), with marking the close and layering 9 each (4%). It has no Ponzi scheme records, and 14 (7%) carry spoofing or layering. SEBI is led by misleading issuer disclosure (176, 27%), price manipulation (157, 24%), matched orders (120, 18%) and insider trading (101, 15%), with wash trading (54, 8%) and front running (53, 8%) behind them; the tags overlap, and 85 of its records (13%) carry no technique tag, many of them issuer or broker fraud that the library’s technique list does not describe. It has no Ponzi scheme or unregistered-distribution records, and the library left out unregistered investment advisers and the NSEL matter by design, so that absence says nothing about what SEBI brings. The FCA is the benchmark and FX agency: 22 of its 48 records (46%) carry benchmark submission rigging or FX fixing, none of which appears in ASIC, OSC, AMF, SEBI or SEC records, and the CFTC has 18 (5%). Records with no technique tag: 14% at the SEC (263), 14% at the CFTC (48), 5% at ASIC (6), 15% at the FCA (7), 33% (23 of 69) at the OSC, the largest share of any agency, 1% (2) at the AMF and 13% (85) at SEBI.

Technique tags are this library’s, not the regulators’, and some are judgement calls (the audit post lists them). The mix differs between agencies partly because the agencies oversee different markets (the CFTC derivatives, the SEC securities) and partly because of what each source publishes; the data cannot say how much of each. The misleading-issuer-disclosure tag was added for the AMF and SEBI records and applies to no SEC record, although the SEC brings accounting and disclosure cases; the SEC rows are filtered by a classifier built for manipulation, not for disclosure.

Outcome. The recorded status reflects what each regulator’s release reports, and the SEC and CFTC use their documents differently from the others. 147 SEC records (8%) and 30 CFTC records (9%) are still recorded as filed, meaning the library has not tracked an outcome; 12 SEC and 2 CFTC records of unknown outcome are separate. ASIC has 3 filed records. The FCA publishes final notices, which are decisions, so none of its records is “filed”, and the OSC, AMF and SEBI have none either. For the AMF, 167 of 209 records (80%) are sanctions decisions, 24 (11%) homologated settlements and 18 (9%) dismissals. ASIC’s media releases often announce a charge and, later, a sentence; 87 of its 96 judgment records carry the criminal flag. Its 7 settled records are too few to compare. SEBI’s 656 are 466 orders recording a finding or direction (71%), 68 settlement orders (10%) and 122 dismissals (19%).

Money. Restricting to resolved records (settled or judgment), 60% of the SEC’s have a penalty stored (1,000 of 1,663), 81% of the CFTC’s (244 of 302), 34% of ASIC’s (35 of 103), 48% of the OSC’s (31 of 64), 67% of the FCA’s (28 of 42), 99% of the AMF’s (189 of 191) and 87% of SEBI’s (462 of 534). The project’s ingest notes record that CFTC headlines routinely name the money, which may help explain the CFTC’s high rate. 60 ASIC records, 53%, carry a prison sentence length, and 40 of them report no civil figure, which may help explain the low ASIC rate. The SEC resolves most matters with a combination of penalty, disgorgement and interest, and 53% of SEC records have a disgorgement figure, against 20% of the CFTC’s. The AMF’s two judgment records with no stored fine are the exception to a pattern in which a sanctions decision nearly always states one. SEBI’s 69 judgment records without a stored penalty are of three kinds: 5 are the grouped batches described below, 40 record a restraint from the market, a suspension or a similar sanction (I did not check each for a rupee penalty), and 24 are other orders, 19 of whose summaries mention a penalty or rupee sum (a keyword check, not a reading of each) that the record does not total, which is a gap and not a zero penalty. Some orders also direct disgorgement that the library does not store.

Bans. The barsImposed field holds each regulator’s own vocabulary: SEC officer-and-director, penny stock and industry bars, CFTC and OSC trading bans, FCA prohibition orders, ASIC bans from financial services and SEBI restraints from the securities market for a set period, suspensions of a registration and, in at least one record, a warning. A “ban” is therefore a different restriction at each agency, and the shares are not like for like. Excluding 65 SEC records and 4 CFTC records whose only entry is a conduct-based injunction (not a bar from an activity), the SEC rate is 28% and the CFTC’s 28%. The OSC’s 25 records with a ban (36%) are the highest share, SEBI’s 164 (25%) are a little below the SEC’s and CFTC’s, and the FCA’s 3 and ASIC’s 10 are too few to say more than that bans are uncommon in the records of those two. The AMF’s 4 (2%) are professional bans on individuals, one for 2 months and two for 10 years, and one permanent ban on an adviser.

Criminal case alongside. A criminal parallel is recorded for 583 SEC records (32%), 108 CFTC (32%), 13 OSC (19%), 8 FCA (17%), 6 AMF (3%) and 1 SEBI (0.2%). The AMF’s 6 (3%) and SEBI’s 1 are the lowest: their records are administrative decisions, and a criminal case running alongside appears only where a decision mentions one. SEBI is the plainer case. It penalises and restrains and does not prosecute, 245 of its record narratives state that the order describes no criminal case, and the single flagged record is an order that notes civil and criminal recovery steps had begun. A false flag on a SEBI record means “the order does not report one”, which is a statement about the order, and nothing about whether anyone was prosecuted. ASIC stands apart at 92 of 113 (81%), and the flag may not mean the same thing there. For the SEC and CFTC it usually marks a Justice Department prosecution of the same conduct, announced alongside a civil action. In ASIC’s records, 91 of the 92 flagged records describe a criminal charge, plea, conviction or sentence in their summaries (a keyword check, not a reading of each), and 60 carry a prison sentence length, so the flag marks that the record concerns a criminal case, and not necessarily a second action running beside a civil one. Four unflagged ASIC records also use criminal vocabulary, which I have not checked individually. The ASIC share therefore says something about what ASIC’s releases announce, and cannot be set against the SEC’s as a measure of how often the two agencies pursue conduct criminally. The 4 October corrections raised ASIC’s share (from 72% to 81%) rather than lowering it, so the earlier explanation still stands.

SEBI: what it publishes, and how its records differ

SEBI is India’s securities regulator. It is the seventh in the library and the one whose records differ most from the other six, so they get their own section. Nothing here is Indian law; it is what 656 records, each written from one AI reading of an order, say.

What SEBI publishes. SEBI does not announce cases the way the SEC or CFTC do, in press releases. It publishes the orders themselves, in English, on an enforcement-orders listing with the order as a PDF: final orders of whole-time members, adjudication orders of adjudicating officers, settlement orders and interim orders. The library read 4,624 listed orders from 1 January 2022 to October 2026 and kept 656 as records. By the record’s court field, 389 were issued by an adjudicating officer, 148 by a whole-time member or members, 74 by an executive director or chief general manager, and 45 do not say. The scope is wider than the manipulation cases of the other six regulators. Following the library’s precedent of including related conduct, it covers fraud on investors or the market under the PFUTP Regulations and section 12A of the SEBI Act: market manipulation, insider trading and front running, but also issuer fraud (such as the misuse of proceeds from a share or depositary-receipt issue), diversion of funds with fraud-regulation penalties, broker fraud on clients with fraud-regulation findings, and fraud matters in which the allegation was dismissed. For many of those there is no matching technique in the library’s list, which is why 85 records (13%) have none. Rejected, and so not in the library: compliance-only orders (listing, takeover and insider-trading disclosure rules, codes of conduct, inspections, registration, summonses), orders where the fraud charge was held not made out and only compliance penalties remain, interim and ex-parte orders (provisional, with no finding), procedural and consequential orders, orders on unregistered investment advisers, the NSEL matter, and mis-selling cases without fraud-regulation findings. The first selection kept 489; 196 records that had been held out or rejected early (189 held-out records and 7 earlier rejects) were read later, and 167 of them were added on 9 October (29 rejected), and about 100 other orders read in the first round were rejected. So 656 is a selection of what SEBI published, chosen by this library’s rules, and not a count of SEBI’s enforcement. The documents behind the records are SEBI’s own, which makes them richer than a press release; the library’s terms for reusing them are to link and summarise in its own words, which the pages do.

Volume, 2022 to 2026. The library holds 312 SEBI records from 2022, 153 from 2023, 68 from 2024, 75 from 2025 and 48 from 2026 (to 29 September). The SEC’s records for the same years are 136, 160, 157, 94 and 73. The SEBI series falls from a large first year, and the year is the date of the order, not of the trading. In many of these orders the conduct is years earlier: one 2022 order concerns trading in 2006 and 2007, and the 2026 batch of option orders concerns trades in 2014 and 2015. Because the library’s window opens in January 2022, the high 2022 count may reflect older cases being decided as much as anything about 2022; the data cannot say.

The status mix, with 122 dismissed. Of the 656, 466 are recorded as judgment, 68 as settled and 122 as dismissed. The dismissals are orders that disposed of a case without a penalty: 88 by adjudicating officers, 24 by whole-time members and 10 by executive directors or chief general managers. Dismissals were 18% of the first 489 records and are 19% of all 656, so the added issuer and broker fraud matters barely changed the share. In the eight I read (from the first selection), seven closed the case because the officer found the allegation not proved, and one because a tribunal and the Supreme Court had already rejected SEBI’s case on materially identical facts. Examples are a trader whose 135 self-trades the officer found too small and unconnected to show manipulative intent, and a case against 16 noticees over trading in an IT company’s shares around its results where the order held the evidence did not show unpublished results were passed on. 68 of the 122 are from 2022. These are in the library on purpose: a record of a regulator’s case that failed is part of the record. It also means “SEBI orders in the library” should not be read as “violations found”; only 466 are recorded as judgments and 68 as settlements, and with the wider scope a judgment may be a fraud finding against an issuer or a broker rather than a manipulation finding. A judgment is also not always against all the noticees in an order: 492 of the 656 records (75%) name more than one respondent, and some orders penalise some and clear others. 57 records mention a remand in their summary or narrative (a keyword count); in two I read, the Securities Appellate Tribunal had sent a case back for a fresh decision, so some orders re-decide a case an appeal returned.

Money, in rupees and kept apart. 462 SEBI records carry a stored penalty, in INR. The 25th percentile is INR 1,712,500, the median INR 4,500,000 (Rs 45 lakh) and the 75th percentile INR 14,875,000. The largest is the 2024 settlement in which the National Stock Exchange of India paid INR 6,430,530,000 (Rs 643 crore) for itself and nine officials, 23% of the INR 27.98 billion (Rs 2,798 crore) the records add up to; the second is the order on Reliance Home Finance, Anil D. Ambani and others (12 noticees), INR 6,240,600,000 (Rs 624 crore). Without the largest the median is unchanged. Two points about these numbers. First, they include sums of multi-noticee tables made by the reading agents: 75% of records name several respondents and the stored figure is the total, so a record’s amount is not what any one person was fined, and it is sensitive to how an agent added a table. Second, SEBI penalties repeat because adjudicating officers impose round amounts: 63 amounts recur across 306 records, led by Rs 10 lakh on 30 records and Rs 5 lakh on 28. I did not treat these as duplicates. Records that name an entity have a median of INR 5,500,000 (288) against INR 2,700,000 for individuals only (173), the same direction as at the other agencies. Settlements have a higher median, INR 10,487,903 (65 records), than judgments, INR 3,200,000 (397); that is the reverse of the CFTC and OSC pattern, and the settlement group includes records that sum amounts across thousands of applicants (see below) and the exchange settlement above, so it is not a clean comparison.

The “illiquid stock options” batch, as five grouped records. From 2022 SEBI’s adjudicating officers issued about 1,580 near-identical orders, one per noticee, penalising reversal trades in illiquid BSE stock options in 2014 and 2015, typically INR 500,000 each. Writing one record per order would have made one conduct, the largest group in SEBI’s output, dominate every count. The library instead holds five records, one per year, that each group a batch: 368 orders in 2022, 860 in 2023, 61 in 2024, 136 in 2025 and 152 in 2026 (1,577 in all). Each is read from a sample of its orders, not all of them, says so, and carries no stored penalty. So SEBI’s 656 records stand for far more than 656 orders; a count of records is neither a count of orders nor of noticees. Two further records, a 2023 settlement scheme covering 10,980 settling entities (INR 1.19 billion, Rs 118.8 crore) and a 2024 one covering 768 applicants (INR 108 million, Rs 10.8 crore), sum the amounts paid; they are among the largest settled figures in the SEBI data and should not be read as single settlements.

How SEBI’s records contrast with the other six.

• Source. The records rest on the full order, where the SEC and CFTC records mostly rest on a press or litigation release and the FCA’s and AMF’s on a decision. That gives SEBI records more reasoning and more detail on how a finding was reached, and more room for the reading agent’s summaries to be wrong.
• Dismissals. 19% of SEBI records are dismissals, against 9% at the AMF and 1% at the SEC. Much of the difference is how the sources work: SEBI publishes the orders that clear people as well as those that penalise them, and the library kept them.
• Respondents. 75% of SEBI records name more than one respondent (SEC 44%, CFTC 52%, AMF 71%) and 62% name an entity (SEC 37%, CFTC 66%, AMF 62%).
• Conduct. 35% of SEBI records touch the order-book family (SEC 5%, CFTC 44%) and 32% the information family, led by misleading issuer disclosure, price manipulation and matched orders; 13% carry no technique tag. There is no Ponzi group.
• Criminal case. One record in 656 carries the flag, for the reason given above, against about a third at the SEC and CFTC.
• Money. Rupee penalties, a median of INR 4.5 million, no disgorgement field and some summed tables; none of it can be set against dollar, euro or sterling figures.
• Reading. The AMF’s 209 records had a sampled second read on 7 October: 54 of 60 agreed on every field (90%, interval 80% to 95%) and 59 of 60 on the core fields. SEBI’s first 489 records had one reading and a sampled second read of 60 (56 agree on every field, 93%); the 167 added on 9 October have one reading and no second read. Each is one AI reading of one order, with no lawyer involved, so the error rate of the later group is unmeasured and the AMF’s 90% does not carry over to either.

Penalty size, in each currency

Every row below is in its own currency. Do not read across a row.

Within the US dollar, the CFTC median is about nine times the SEC’s (US$1.5 million to US$173,436, a ratio of 8.6), and the CFTC’s smallest stored penalty is US$17,500 where the SEC’s is US$1,000. The two agencies’ records also differ in who they are about: 66% of CFTC records name at least one entity, against 37% of the SEC’s. Restricted to records naming only individuals, the medians are US$113,444 for the SEC (573 records) and US$350,000 for the CFTC (66); where an entity is named, US$375,000 (432) and US$2,134,083 (176). The same pattern holds in each other regime where the group is large enough: OSC records naming only individuals have a median of C$76,663 (20) against C$650,000 where an entity is named (12). ASIC splits the same way (A$42,840 for 25 records and A$832,500 for 9), but the second group is below 10 records and should not be treated as a comparison. The AMF shows the same split: 70 individual-only records have a median of €180,000 against €545,000 for the 119 that name an entity. SEBI shows it more weakly, INR 2,700,000 for 173 individual-only records against INR 5,500,000 for 288 that name an entity. The comparison within each agency is between kinds of respondent. It is not between regulators.

The FCA is two populations. Of its 30 penalty records, 12 name an entity, with a median of £216,681,500 and a minimum of £630,000; all 12 were filed in 2013 to 2015, and most are the benchmark-rigging and FX cases against banks, with a broker and one layering firm. The 18 naming only individuals have a median of £97,800. The £279,922 overall median falls in the gap between the two and is representative of neither. The 75th percentile of £100,625,000 shows it. With 30 records the figures are sensitive to a few cases.

Outliers. One record dominates each total, and the medians barely move when it is removed.

Totals are shown only to put the outlier in proportion; they are sums of records, not of regulators’ total penalties, and several records in each group are the same matter at different stages. The SEC total without Stanford is US$1.89 billion. Two ASIC records hold the A$5 million (the CBA matter at two stages), so the table’s removal leaves another A$5 million in the data, and the ASIC median barely moves. The CFTC’s largest record is a joint order against five banks, stored as a sum across the respondents, and SEBI’s is a settlement payment by an exchange for itself and nine officials (the next largest, Reliance Home Finance and others, is a sum across 12 noticees).

Settled against judgment. Within the SEC, the median settled penalty is US$129,140 (687 records) and the judgment median is US$434,887 (313). Within the CFTC, US$1,000,000 for settled records (69) and US$1,600,000 for judgments (175). The OSC’s settled median is C$200,000 (21) against C$450,000 for judgments (10). The AMF’s homologated settlements have a median of €100,000 (24) against €420,000 for sanctions decisions (165); the settlements are a small group and may concern different conduct. SEBI’s run the other way, INR 10,487,903 (65) settled against INR 3,200,000 (397) for judgments, for the reason given in the SEBI section. The ASIC and FCA groups are too small to compare: ASIC has 6 settled and 29 judgment records with a penalty, the FCA 25 settled and 3 judgment. Restricting to matters filed from 2014 to 2025, which is the span the six agencies other than SEBI cover in full (SEBI starts in 2022), leaves the medians unchanged for ASIC (A$65,000) and the OSC (C$267,500), and moves the SEC’s to US$177,139, the CFTC’s to US$1,542,916, the FCA’s to £250,000 and the AMF’s to €475,000 (94 records); the AMF’s 94 records with a fine from before 2014 have a median of €300,000, so its stored fines are larger in the later period.

What this does not show

It does not compare the regulators. These are the cases each regulator chose to bring and publish, passed through this library’s classifier and sources. A low count for an agency can mean the agency brings few such cases, or that the library collects it thinly. The FCA’s 48 records, for example, are final notices that survive a filter that skips most of the FCA’s docket, the OSC’s are Ontario only, and SEBI’s 656 are a selection from 4,624 listed orders. Which matters an agency pursues, and how it chooses to resolve them, shape what is published. The records cannot show how much manipulation each agency faces, how it chooses cases, or how severe its penalties are, and a larger penalty in a record does not mean a stricter regime.

It does not compare penalty sizes across currencies. The medians are in US dollars, Australian dollars, Canadian dollars, pounds, euros and rupees, over different years. The table does not support a statement that one regulator’s typical penalty is larger than another’s across currencies, and no exchange rate has been applied.

The data have gaps. 567 SEC records (31%), 79 CFTC (24%), 78 ASIC (69%), 37 OSC (54%), 17 FCA (35%), 20 AMF (10%) and 194 SEBI (30%) carry no money field of any kind, and for ASIC 67 of those 78 are judgments and 3 are filed; SEBI’s are 122 dismissals, 69 judgments and 3 settlements. A missing figure is not a zero penalty. 147 SEC and 30 CFTC records are still recorded as filed, so their outcome is not tracked, and some resolved records cover only some of their defendants. Multi-respondent records store the sum of the respondents’ amounts, which inflates what an agency’s “typical” record looks like. Ten ASIC amounts recur across two or three records (21 in all); in nine of the ten the same person is named on each, which fits one matter at charge and at sentence, and in one (A$810,000) two different people, which I did not investigate. At the AMF 107 records share a fine with at least one other (€100,000 on nine records, €300,000 on seven); round amounts recur across different matters, and some decisions store a sum across several respondents, so I did not treat these as duplicates and did not check them. The same holds at SEBI (306 records share one of 63 amounts). Dates are mostly filing dates, for the SEC many are the day an order was instituted and settled at once, and for SEBI they are the date of the order.

The native-currency fields. Checking every non-US record for money stored in the wrong field now finds one exception: the FCA’s Coscia record carries a US$903,176 penalty with no native amount, which is correct because the FCA’s notice states it in dollars, so it is the one FCA record outside the sterling distribution. It is not counted in the 30 above. No other ASIC, OSC, FCA, AMF or SEBI record has a value in a USD field, no USD amount was found in a native field, and no SEC or CFTC record has a native amount. Separately, 45 resolved ASIC, FCA, OSC and SEBI records (none at the AMF) have an amount in their summary and no stored figure; the count includes the Coscia record, and the pattern also matches losses or costs, so I did not check each; some may be missed penalties.

Checked by AI agents, not lawyers. All records of the first 2,428 were read against the regulators’ own documents on 2 and 3 October 2026 by Claude AI agents under written instructions, and no lawyer reviewed them. The AMF records, added later, were read in French by AI agents in the same way, and 60 of them were re-read by a second agent (90% agreement), so translation and the reading of French sanctions law are further places for error. The SEBI records were each written from one AI reading of one order. The first 489 had a sampled second read of 60 on 8 October that agreed on every field for 56 (93%, interval 84% to 97%; the four misses were all penalty amounts); the 167 added on 9 October have no second read. Some of their figures are sums the agents made. The audit post sets out the method and its limits, and the corrections page logs what it changed. Where a record here described a matter that the regulator only alleged, the figure is the regulator’s claim, not a finding. Each record links to its primary document, and an error link on each case page goes to the maintainer. A record mentioned here is not a verdict on anyone.

An example of the difference in kind: the ASIC release about Westpac’s pre-hedging records a A$1.8 million penalty for unconscionable conduct, and the record carries no technique tag, while the FCA’s Coscia final notice records a layering finding in dollars. The records share a field name and not a meaning.

Techniques referenced

• Insider trading — Insider trading is dealing in securities on material non-public information in breach of a duty of trust or confidence, which exploits a true signal others cannot see rather than injecting a false one into the price.
• Ponzi schemes — A Ponzi scheme pays returns to earlier investors from later investors' money rather than from genuine profits, which is investment fraud with no market involvement and no price to manipulate.
• Price manipulation — Price manipulation is trading intended to create an artificial price — one that does not reflect genuine supply and demand — and it is the general charge a regulator brings when the conduct does not fit a more specific named technique.
• Spoofing — Spoofing is placing orders a trader intends to cancel before execution, to create a false impression of supply or demand and move the price toward a smaller genuine order on the other side.
• Benchmark submission rigging — Benchmark submission rigging is skewing the figures a panel member reports into a reference rate calculation, so that the published benchmark moves in favour of positions that settle against it.

Cases referenced

Reviewed October 7, 2026. Spotted an error? Tell us.