AI tools improve access to derivatives transparency but they don’t close the gap between US and European regulation.

I write extensively about derivatives transparency. Where does it stand in 2026?

US SDR Data

The “gold-standard”, US SDR data continues to deliver:

• Transaction-level derivatives transparency.
• Most of the primary economic terms of trades, but trade descriptions are now driven by “Unique Product Identifiers“.
• Notional sizes capped at reporting thresholds. Full size never disclosed.
• Most trades, particularly in Rates, published within 1-2 minutes of execution. At least 75% of trades are reported within 15 minutes.
• No product limitations – from SOFR swaps to HUF Swaptions, everything is reported. Cleared or bilateral, vanilla or exotic, linear and options.
• Most asset classes are covered, with excellent data in Rates, Credit and FX options/NDFs.
• Package trades reported with a package identifier and price, alongside full leg-level detail.
• Venue identifiers for trades executed on-SEF or equivalent MTFs, and off-venue trades clearly identified as such.
• No counterparty information – all trades are anonymous.
• No direction information.
• All trades reported at the 15-minute threshold – no complicated deferral regimes.

European APA Data

• Transaction-level derivatives data published via ~20+ Approved Publication Arrangements.
• Time to publication ranges from near real-time (<1–5 minutes) to deferrals of EOD, T+1 or T+2 depending on liquidity classification.
• Large trades subject to LIS/SSTI regimes – notionals masked or published in size bands.
• Trade identifiers are published, not economics. APA reports must therefore be combined with reference data.
• Packages reported as individual legs without a common package identifier.
• Venue of execution flagged – including Multilateral Trading Facilities, OTFs, SIs and off-venue trades.
• Instrument coverage includes Interest Rates, Credit and FX derivatives, as well as Fixed Income instruments.
• No counterparty identification in public data.
• No direction (buyer/seller) disclosed.

Underwhelmed?

European data doesn’t meet our “gold-standard” (neither do other jurisdictions – see Appendix below). However, data providers are now able to use European data, as evidenced by ISDA’s work on transparency. Swapsinfo provides transaction data sourced from both APAs and Transaction Venues, summarising weekly notional volumes and trade counts by tenor, product type and currency:

This is a helpful resource and motivated me to go back to the source, with the help of AI tools.

Transparency Data in an AI World

For US data, I have found that I do not need to use AI tools to access post-trade transparency. The DTCC cumulative files for Rates trades are available here, and provide a trade-by-trade view of activity – pretty much straight out of the box:

Anyone familiar with European transparency will know that it is not as straightforward on this side of the Atlantic. But, thanks to AI tools, I knew I was only a prompt away from dealing with the following list of Tradeweb slice files:

A consolidated view was provided via this Gemini prompt:

Write me a simple python script that will download all of these slices csv files and consolidate into a single csv

http://a91bb989ce98.mifid.io.tradeweb.com/html/?prefix=Tradeweb_BV_APA_Post-Trade_TWEA/

The resulting 100-line python script provided me with a consolidated file for 20th April activity. A mighty step forward for European transparency, thanks to AI!

Tradeweb are even kind enough to publish their 600 words of terms and conditions on every single slice file. Every 2 minutes we can read how much they want to charge us for their data. Lucky us!

Fortunately, the Tradeweb file also includes trades:

• ~10,000 transactions.
• As expected, the primary economic terms of the trades are not displayed.
• The file doesn’t include any asset class information.
• This is no fault of Tradeweb. The rules only require the ISIN or equivalent instrument ID to be published.

AI Phase Two

Undeterred, I turn to Gemini again:

Here is the consolidated file. I need to decipher Column E, the instrument identification code. Write a python script or suggest a plan to access the primary economic terms of each instrument. I need asset class, currency, start date, end date, product type, index.

For most users, most of the time, this step will be fruitful. I followed Gemini’s advice to use OpenFIGI and this resolved 87% of transactions at Tradeweb. That sounds pretty good right? Ah, there is a BUT:

• The missing 13% were mainly derivatives.
• That’s a pain given I only really care about the derivatives data!
• I then spent an hour with Gemini trying to work out how to get the ESMA database API to work (it is called FIRDS – ping me if you want a working python script!).

But, hey presto. A couple of hours with AI and I was able to have a view of European transparency. Okay, it is only one APA, but I think it is the largest.

AI Phases 3 to “n”

And then my fun really began. I wanted to know some basic facts:

• How many trades were reported in EUR OIS in Europe compared to the US?
• How did trade sizes compare?
• Were there any duplicates?
• Do average trade sizes vary across jurisdictions?

To get there I had to do some major data manipulations:

• Merge Tradeweb reported data with ISIN reference data to record primary economic terms.
• Create a cross-jurisdiction schema to create a “golden source” across both Tradeweb and DTCC.
• Ensure only one version of each trade was present – either from amendments or from duplicates across jurisdictions.
• Convert to USD notional equivalents.

This ended up as a deeply involved project, resulting in multiple python scripts doing small parts of the merging and data cleansing:

So here is a health warning – don’t start unless you have a spare day to burn!

Transparency Outputs

And what did I find? Working with AI tools pretty much gives you a Streamlit dashboard for “free” nowadays, so I can show you my findings:

• US SDR data is still four weeks ahead of EU APA data.
• It is a neat visualisation of where we still are – those ~20 trading days in Europe are still “dark”, whereas we could easily collect the US SDR data for those intervening days from the DTCC cumulative files.
• Sizes reported to Tradeweb across EUR and USD OIS and IRS appear to be similar in size….
• But average trade sizes are larger on the Tradeweb APA.
• I found no duplicates….because my datasets were effectively 4-weeks apart thanks to post-trade deferral rules in Europe. Annoying that I did all of the hard work before I realised this!

To resolve this I returned to DTCC and grabbed the data for March 19th, matching the trade data of most of the deferred Tradeweb transactions. Apart from an end-of-day peak at Tradeweb, we see that far more of the EUR swaps market is being reported to US SDRs rather than European APAs:

In Summary

• AI tools have improved access to post-trade derivatives transparency. Good news!
• Trades are published with identifiers rather than primary economic terms. Published files cannot therefore be analysed by AI without painful data merges.
• Regulations ensure that there is value in standardising and augmenting published data.
• The US continues to be the gold-standard in terms of:
  • Timeliness
  • Completeness of data
  • Coverage
• A simple search for duplicate trades across US SDRs and EU APAs does not reveal any potential matches. One to dig into further.
• Coverage of the EUR swaps market appears to be more complete in US SDR data than European APA data from this initial sample.

---

Appendix – Other Jurisdictions

FCA UK Data

• As above for MIFID data.
• As of December 2025
  • Simplified deferrals: real-time, EOD or T+1.
  • LIS/SSTI calibrated using UK data.
  • Removed “Traded on a Trading Venue” to determine publication. Publication requirement determined via liquidity assessments per asset class.
• One to add to see if EUR swaps coverage improves given so much swaps liquidity is in London.

Canadian SDR Data

• As per the US above, but T+1 publication.
• Lower notional reporting thresholds.

Japan

• JPY IRS transparency is venue-level – not whole-market SDR-style data.
• Public transaction data comes from trading venues / brokers such as Tradition Nihon – not a single national source.
• Coverage is partial and typically T+1 – reflecting venue flow, not total market activity.