From the courts' own feeds

Who was sued in federal court on Friday, 25 September 2026

711 new federal civil cases reached our feed with this filing date, from 58 district courts. Below are the organisations named as defendant in five or more of them. This day is still running, so every number on the page is a floor and will rise.

The largest cluster

Federal Express Corporation

Named as defendant in 89 separate new federal civil cases, across 5 districts: C.D. Cal. 40, E.D. Tex. 33, S.D. Ind. 10, W.D. Tex. 5, D.R.I. 1. The feed carried 711 new civil cases in total for this date.

Why these were filed together. These are individual re-filings, and they are not new. On 18 May 2026 the Western District of Pennsylvania held that 14,296 delivery drivers suing over unpaid overtime in three consolidated actions — Brannon, Abner and Smith — had been misjoined, severed their claims, and tolled the limitations period so each of them could file a separate suit in an appropriate district. The cases reaching the feed are those separate suits, which is why plaintiff surnames tend to arrive in alphabetical order: a list is being worked through. A day of them is an instalment of a re-filing that order set in motion months ago, not the start of anything.

Source: Courthouse News: 14,296 FedEx drivers’ overtime claims severed. We did not discover this cluster's cause; we counted its filings and then went and found the order behind them.

Everything else filed five or more times

Defendant, as the court wrote itNew casesDistricts
Becton Dickinson and Company56D. Ariz. 56
Commissioner of Social Security12D.N.J. 3, S.D.N.Y. 2, W.D. Wash. 2, C.D. Cal. 1, E.D.N.Y 1, S.D. Ohio 1, W.D. Tenn. 1, W.D. Va. 1
Meta Platforms, Inc.6N.D. Cal. 6
WARDEN, DELANEY HALL DETENTION FACILITY6D.N.J. 6
3M COMPANY5D.S.C. 5

Ranked by count, with nothing added and nothing removed. Several of these will have ordinary explanations — a court that direct-files into a multidistrict proceeding collects suits against one company by design, and a severance order can produce hundreds of separate cases in a week. A cluster is a thing to look into, not a conclusion.

Get this as it happens, for a company you care about

This page is one day's arithmetic, recounted after the fact. A filing stream is the same query running continuously: name a company, a district or a phrase, and every new federal case matching it arrives as the court publishes it, with a morning digest of what came in overnight. The same record is readable inside ChatGPT, Claude, Gemini and Grok, so the case file follows you into whichever assistant you already work in.

Start a filing stream

7 days free, $0 today, then $299/month. Cancel any time.

Connect it to your assistant first or compare the plans.

How this was counted

Every number comes from our own copy of the district courts' CM/ECF filing feeds. Each federal district runs CM/ECF, and most publish a public feed of new docket activity that sits outside the PACER paywall. Our poller reads those feeds directly and stores what they say. Nothing on this page passes through any intermediary, and no part of it is generated by a language model.

A new case is a docket entry numbered 1, or one the court labelled as a complaint, a notice of removal, a petition or a civil cover sheet — the filings that open a federal civil case. Cases are counted once each, never by entry: a district that publishes four separate notices when a case opens contributes one case, not four.

A cluster is an organisation named on the defendant side of five or more of those captions on the same day. Defendants are grouped by what the court wrote, after normalising capitals, punctuation and the spelling of legal suffixes — so Equifax Information Services LLC and Equifax Information Services, LLC are one row. Nothing further is merged: two differently worded names stay two rows, even when we suspect the same company, because a row a reader cannot check against the docket is worse than a row we did not combine.

The day is the court's day, taken from the filing date the court itself published, not from when our poller happened to read the feed. That matters: our poller has been down and caught up in a burst, and a page that bucketed on our own reading time would have reported the catch-up as a surge of filings. A day runs midnight to midnight US Eastern, so a court in the Pacific time zone filing late in the evening lands on the following day here.

Defendants named only as individuals are never shown, however many times they appear. The test is textual and deliberately cautious: a caption has to carry a word that names a legal entity or a public office. It excludes real companies whose names do not say what they are, and that is the direction we would rather be wrong in. On this date 287 of the 711 new cases were left out of the grouping on that test, or because the caption had no readable defendant side at all.

What this is not

It is not a census. Fifty-five of the ninety-four federal district courts publish a complete feed, twenty-six publish only certain entry types and thirteen publish nothing at all, so a case can be filed and never appear here. Every district, and what each one publishes. Counts are a floor.

It is not an accusation. Being named as a defendant is the start of a case, not the end of one. Nothing here states or implies that any claim has merit, and no count should be read as evidence about any company's conduct.

It is not about the people who filed. Plaintiffs are not named or counted individually anywhere on this page. Only the organisation being sued, the district and the count are published.

It is not a docket. A feed item carries the court's one-line label for what was filed and never the filing itself, so nothing here reports what any complaint says.

Some entries from four districts — W.D. Tex., C.D. Ill., D. Conn. and S.D. Tex. — reach us carrying a filing date in the year 3926. That arithmetic error is in the courts' own feeds rather than in our reading of them: our parser reproduces the published value exactly, and almost every affected entry is one auto-generated docket label. We do not repair it, because the month is wrong as well as the year and a repaired date would look exactly like a date the court gave us. We hold no filing date for those entries at all, and an entry with no filing date belongs to no day, so it is left out rather than quietly counted somewhere. A case whose only opening entry has no usable date is therefore missing from this page. No date on this page comes from anywhere but the court. On this date every case we saw open carried a date we could use.

Sources and credit

The filings are the courts' own. The data on this page is read directly from each district court's public CM/ECF feed by our own poller — it does not come from any third-party docket service, and no licence or API key sits between these numbers and the courts that published them.

The catalogue of which district publishes a feed, where it lives and which of them are irregular follows Juriscraper, maintained by the Free Law Project, whose work is why anyone can read a federal docket without paying for it. We are grateful for it, and it is the feed addresses we take from them, not the data.