Start with the date, because it is the news. On August 5, CDC’s investigation page said illnesses linked to this outbreak began on June 22 and continued through July 31. On August 13, the same page says they began on June 14 and continued through August 3. The window opened eight days at the front and three at the back. The back end is ordinary, because cases keep arriving. The front end is not ordinary at all. It means that as the interviews caught up, this investigation absorbed people who got sick more than a week before anybody previously placed the beginning of it.

June 14 matters because of what the recall actually reaches. FDA’s advisory says distribution of the recallediceberg ran from June 29 through July 16, and the recall notice covers product made in that window. Symptoms of cyclosporiasis begin about a week after exposure, with a range of two days to two weeks, so the person who became ill on June 14 was most likely at a table around June 7. That is three weeks in front of the earliest lettuce in the recall. I wrote on August 8, working from FDA’s own enforcement file, that the recall window was set by shelf life rather than by contamination, because you cannot recall lettuce that has already been eaten or expired. The federal outbreak’s own start date has now moved out in front of the recall by more than two weeks.

FDA published once in seven days, on Thursday night, and when it did the outbreak grew by half. It went from 6,358 illnesses, 278 hospitalizations and fifteen states on August 5 to 9,481 illnesses, 398 hospitalizations and seventeen states on August 13. Maine and Massachusetts joined. Confirmed distribution went from twenty-seven states to thirty-one, with nine more named as possible, and that finally closed the West Virginia contradiction the agency had carried for weeks, in which a state was on the outbreak list and missing from the distribution list at the same time. FDA also said two things it had never said before. Inspectors are inside the plant at Doctor Mora alongside Mexican officials, and customer records from the firm show recalled product was sold to consumers in Mexico.

CDC’s surveillance page went from 10,468 laboratory-confirmed domestic cases on August 4 to 13,895 on August 11, up roughly a third in a week. Hospitalizations went from 517 to 740, up 43 percent. Travel-associated cases, which are counted separately and are not this outbreak, went from 1,341 to 1,530, and their hospitalizations from 46 to 58.

The states went from 26,755 on August 7 to 30,513 tonight, and past thirty thousand. Of that, 3,758 is seven days of growth and 10 is a correction: Mississippi and South Dakota turn out to have published counts I had been carrying as blank, 7 and 3, dated July 22 and July 16. Those are not new illnesses; they are rows that were missing. Michigan alone was 1,424 of that, moving from 12,485 to 13,909 in its single Thursday update. Ohioadded 552 and crossed five thousand, with Cuyahoga County at 695 now ahead of Lucas at 690, so the center of that state’s outbreak has shifted from Toledo to Cleveland. Missouri added 482, Kansas 287, Indiana 278 to reach 1,983, with 486 of them in Allen County alone, Florida 153, New York 145 and Nebraska 127. Maineappeared on the chart for the first time at 27. West Virginia posted this morning at 304 and revised its hospitalizations down from 21 to 20, the first time any state on this chart has revised a hospitalization figure downward. 

Where the state governments are is a harder question than where the federal agencies are, because the states are not all in the same week. Fourteen of them added cases over the seven days. One revised down. Thirty-four published no new number at all. Kentucky is dated July 29 and Iowa July 30, both more than two weeks old. Nine states have published nothing since a press roundup on July 17. Illinois, Texas and Massachusetts did not move at all. 

Here is where all fifty of them stand tonight. The last column is whether FDA has confirmed that recalled Taylor Farms de Mexico lettuce reached that state. Thirty-one are confirmed, four more than a week ago, and nine further states and territories are marked as places the product may have gone. Forty-one states now publish a 2026 count of their own, forty of which are added into the total below.

30,523 is a floor, not an estimate. It is the sum of what forty health departments have published, on schedules running from daily to never, counted at the bottom of any range a state reports. Hawaii is the one state that publishes a number I do not add. All five of its cases are travel-associated by its own health department’s account, and this is a domestically acquired figure, which is the same reason CDC’s 1,530 travel cases sit outside it. Nine states publish nothing at all, and one of them, South Carolina, is on the federal distribution list and reported forty-five cases last year, so it plainly does make the disease notifiable. 

The floor is also the smallest of the three numbers in this piece, and none of them is the number of people who got sick. Most people with cyclosporiasis never get a stool test, because the parasite is missed by the routine ova and parasite exam and has to be asked for by name, and because somebody with a diarrheal illness that comes and goes for a month often never sees a doctor at all. The standard estimate of that gap comes from Scallan and colleagues, Foodborne Illness Acquired in the United States, Major Pathogens, in Emerging Infectious Diseases in 2011, which puts the underdiagnosis multiplier for Cyclospora at 83.1. Applied to CDC’s 13,895 laboratory-confirmed cases, that puts the real number near 1,155,000. It is an estimate with wide margins, and it is not a headcount, and I use it because the alternative is to let 30,523 stand as though it were the answer. It is not. It is the part we can count.

The date is the thing I would go looking at. Somebody at CDC changed June 22 to June 14 this week, and whoever did it was doing the job right. It also means this outbreak now officially begins two weeks before the earliest lettuce the recall reaches, and the meal behind that first illness came three weeks before it. Inspectors are finally standing in Guanajuato. The question they went there to answer just got two weeks older.

The fifty state health departments now report 29,920 cases of cyclosporiasis this season, eighty short of thirty thousand. The federal outbreak grew by half this week, and inspectors are finally inside the plant in Guanajuato.

A tip of the pen, because a good deal of the below is not my reporting. CNN’s investigation — Kyung Lah reported it, with Brenda Goodman’s earlier work on the traceback and the false positive, and Katherine Dillinger contributing — produced the inspection history that makes the timeline legible. Rachel Roubein, Lena H. Sun and Caitlin Gilbert at the Washington Post put the interval itself in a headline this week, which is the framing I have borrowed here. CBS News independently reported the seven-year gap. Reuters has carried the Mexican announcements throughout, Sarah Morland on the negative results and Siddhi Mahatole, Mariam Sunny and Mrinalika Roy on the Michigan counts out of Bengaluru. Coral Beach at Food Safety News has tracked the company’s statements. Food Safety Magazine had the inspection announcement first. And Phyllis Entis at eFoodAlert remains the only other person who went through the recall file line by line.

Two other things happened this week and they are being reported as though they happened together. They did not. Mexico’s Secretariat of Health announced that lettuce and water samples from the Taylor Farms de México plant in Doctor Mora, Guanajuato tested negative for Cyclospora cayetanensis. And FDA announced it had begun an onsite inspection and sampling at that same plant. The first happened in July. The second happened on August 13. Three and a half weeks separate them, and the gap is the story.

Take the Mexican findings first, because they are being read as an exoneration and they cannot carry that weight. Cofepris inspected the facility from July 18 to July 20, accompanied by epidemiologists from the Directorate General of Epidemiology and food safety specialists from Senasica. Its National Reference Laboratory and the Guanajuato State Public Health Laboratory ran real-time PCR on ten samples of raw materials and finished products. All ten were negative. Water samples were negative for fecal coliforms and within national physicochemical standards. The epidemiologists looked at worker health and found no diarrheal illness requiring intervention. I have no reason to doubt any of that. I doubt what people are making of it.

The dates answer most of it. Those samples were taken one to three days after the July 17 recall, and roughly four weeks after the median patient in this outbreak got sick. The lettuce that made people sick was cut and packed from late June through the middle of July. You cannot sample a plant on July 18 and learn what was on the leaves that left it in June. A negative result in the third week of July is a fact about the third week of July.

The fecal coliform result deserves its own paragraph, because it is the exact mistake I have been writing about all summer and a national government just made it in an official release. FDA says in its own voice that traditional fecal-indicator testing — generic E. coli, fecal coliforms — will not identify this parasite. Reporting clean coliforms as reassurance about Cyclospora is like testing a swimming pool for chlorine and announcing there are no sharks. Every water-quality standard in the produce industry is built on indicator bacteria, and the organism at the center of this outbreak is invisible to all of them. Ten samples is also a very small number for a contaminant that is sporadic and patchy and present at low levels when it is present at all.

The worker-health finding is the one I would push on hardest, and gently, because it is also the oldest thread here. Mexican epidemiologists found no diarrheal illness among workers in the third week of July, after the implicated harvest was over. In 2013, when a large cyclosporiasis outbreak was traced to this same Doctor Mora facility, FDA’s environmental assessment found that medical records of harvesting workers showed antidiarrheal medications and the antibiotic combination used to treat cyclosporiasis, though no diagnoses were recorded in the records inspectors saw. That proves nothing standing alone, and I say so every time I raise it. What it establishes is that looking for sick workers in July is not the same as knowing who was sick in June, and that somebody has been down this road at this address before. Senasica added one detail that reads as routine and is not: it invited the supplying growers to certify under Mexico’s Contamination Risk Reduction Systems program. You extend that invitation to people who do not already have it.

Now the American half. FDA’s inspectors reached Doctor Mora on August 13. Taylor Farms’ own information hub says the company temporarily suspended production of all products at that facility as of July 18, 2026. The inspection that began this week began at a plant that had been shut for twenty-six days, sampling lines that had not run since the day after the recall. How long had it been before that? CNN’s investigation established the answer. FDA had not sent an inspector to the Taylor Farms Mexico plant since 2019, and before that not since 2013 — the year of the last outbreak traced to the same address. In the first six months of 2026, five of the company’s more than 160,000 produce shipments from Mexico were physically inspected at the border. Taylor Farms’ Mexican operation is the single largest source of vegetable and prepared salad shipments into this country of any manufacturer in the world.

I want to be careful about where the blame for that lands. Not on the investigators. The people who ran the traceback that found this plant did that work in weeks, with a parasite that has no whole genome sequencing and no positive sample to anchor to, and they got there. The CORE analysts, the state epidemiologists and the traceback teams have been the functioning part of this response. Nobody was assigned to be in Doctor Mora in June. 

What would actually answer the question at this point is not another round of product sampling. The product is gone; the plant is idle and the lots that made people sick expired in July. What would answer it is the environmental record: the water sources and their history, the sanitation infrastructure around the growing fields, the worker medical and absence records from May and June rather than July, and the import history showing what crossed the border in the weeks before anyone was looking. FDA produced an environmental assessment at this facility in 2013 and published it. The 2026 version is the document to wait for.

The season closes on August 31, eighteen days from today. Inspectors are in the plant. The counting does not stop either way.

Yesterday I counted twelve letters from members of Congress about this outbreak. It is thirteen. Senator Elizabeth Warren wrote to Health and Human Services Secretary Robert F. Kennedy Jr. on August 13, the same day I published, and her letter is the thirteenth. Five from the House, eight from the Senate. She asks nine questions and wants them answered by August 27.

That is the sixth time this count has moved on me, and it is worth saying plainly that the problem with these letters is not the people writing them. The questions in Warren’s letter are the most specific anybody has asked. Her first question has six subparts and every one of them goes at the July 18 positive test that FDA retracted on July 19. She asks what protocols exist to verify a result before it is released, why FDA reevaluated the finding, whether it reevaluated because Taylor Farms or its representatives asked, whether a political appointee asked, and she demands a transcript of the reported call among the White House, the company and FDA. Nobody else has gone near that sequence in this much detail.

The subpart that matters most is the last one, and it is the only question in thirteen letters that is actually about food safety rather than about politics. If confirmation testing is a standard part of FDA’s procedures, why was the positive communicated to the public before the sample was reevaluated?

Three more of her questions deserve answers whatever anybody thinks of the politics. Her fifth and sixth ask what it costs the department to track, trace and treat cyclosporiasis, broken out for each year from 2016 through 2025, and then what this one has cost. Her seventh asks how many of the states formerly required to report Cyclospora under FoodNet have stopped reporting since July 2025, with a state-by-state breakdown. Every letter since July 14 has asserted that the FoodNet change damaged surveillance, and that question is the one that would prove it or disprove it. She also asks CDC for its own estimate of how many cases have gone unreported, which would be the federal government finally putting a multiplier next to its own count.

Here is the thing about all four of those questions. The answers exist. They are sitting in federal file cabinets and federal spreadsheets right now, and any competent analyst at CDC could produce the FoodNet breakdown in an afternoon. Nothing in Warren’s letter can make anybody walk to the cabinet. Seven dated deadlines across these thirteen letters have now expired and I cannot find a published answer to a single one of them. Senator Ossoff gave the Secretary seventy-two hours, got nothing, and had to spend a second letter saying so. A letter from a member of Congress is a request. It is answered when the agency feels like answering it, and this agency does not feel like it.

Here is the whole list, updated. The column on the right is the one to read.

That brings me to the part I actually want to write about. There is one committee in this fight that can compel an answer, and nobody has asked it to.

Senator Richard Blumenthal wrote to FDA and to Bruce Taylor on July 30 as ranking member of the Senate Permanent Subcommittee on Investigations. PSI is the Senate’s standing investigative body, and it has subpoena power. Those two letters ask for every communication between FDA and the White House about this outbreak and about Taylor Farms, and the records definition names Signal and WhatsApp by name. Both letters were copied to Chairman Ron Johnson. Both deadlines were yesterday. Neither has been answered, and a ranking member cannot issue a subpoena alone. That takes the chairman’s signature or a vote of the subcommittee. One person could change the posture of this entire investigation by close of business and has not.

On the House side, Robert Garcia sent Taylor Fresh Foods five categories of documents and five written questions on July 27, including every communication with the White House since January 2025 and every sanitation, audit and contamination record for the plant in Mexico going back to 2013. The deadline was August 10. No production has been reported, and the company is under no obligation, because Garcia is in the minority and a minority document request is a letter with better stationery. The Oversight majority could adopt that same request this morning and it would be answered, because that request would have the committee behind it.

And there is the oldest tool of all, which is a hearing. Put the commissioner and the director under oath, put the company in the room, and ask the questions on the record with a transcript. This is not a novel idea for this parasite. The same Permanent Subcommittee on Investigations held a hearing on Cyclospora in imported produce in July 1998, with CDC scientists and a woman who had been sick testifying about the same problem. Twenty-eight years later, same subcommittee, same parasite, no hearing.

There is also something Congress can do without anybody’s permission, and not one of the thirteen letters asks for it. Warren asks whether the March 2025 delay of the food traceability rule was influenced by a million dollars that went to MAGA Inc. the following week, and by another hundred thousand that went to a second Trump-aligned committee on June 13, 2025, a few weeks before the administration dropped mandatory reporting of Cyclospora and five other pathogens. That is a fair question, and she is careful to frame it as a question. But asking whether the delay was bought is not the same as ending the delay. Section 780 of last November’s appropriations act bars FDA from spending a dime to enforce that rule before July 2028 and instructs the agency to go looking for ways to loosen the lot-level tracking requirement, which is the part of the rule that would actually let anybody follow a case of lettuce backward. Eleven consumer, public health and labor groups wrote to the four congressional leaders on August 12 asking them to rescind it. Thirteen members have written letters about this outbreak and none of them has asked for that. A rider can be undone by the same body that wrote it, and nobody needs the Secretary’s cooperation to do it.

The letters have also turned up facts that then went nowhere, which is its own kind of failure. Warren’s letter reports that CDC cut the staff responsible for tracking and identifying parasites from eleven people to three. It reports that USDA’s Beltsville Agricultural Research Center is being closed and relocated, and that two of the laboratories inside it investigate and sequence this parasite. That second one ought to have stopped somebody in their tracks. There is no whole genome sequencing available for Cyclospora, and CDC says so on its own page. When CDC decides that a sick person in Ohio and a sick person in Michigan belong to the same outbreak, it is not matching genomes the way it would for Salmonella or E. coli. It is working from genetic markers, and the methods behind that work come out of a very small number of federal parasitology laboratories. Two of them are being packed into boxes while the largest outbreak of this parasite in American history is still going. That fact appears in one footnote of one letter, and nobody has followed it anywhere.

And then there is the question none of the thirteen has asked at all. Not one of these letters asks how the parasite got onto the lettuce. Nothing about irrigation water. Nothing about the sewage and sanitation failures in the communities around the plant at Doctor Mora. Nothing asking why an industry that tests constantly for indicator organisms does not test for this parasite, or whether a validated test for it on finished product even exists. Nothing about the fourteen states that received the recalled product and have no attributed cases. Nothing asking whether consumers in Mexico, where FDA now confirms the recalled lettuce was sold, were ever told. FDA likely got inspectors inside the plant in Guanajuato on August 13, twenty-seven days after the recall, and not one member of Congress had asked when that was going to happen.

That gap matters more than it looks. Taylor Farms de Mexico is a processing plant. The lettuce arriving at it is grown by other people on other ground, with other water. If this parasite came off that water, then everything Congress is asking about, the phone calls and the donations and the press releases, sits downstream of a problem nobody has put in writing to anybody. The counting problem and the corruption problem are both real. They are also both consequences.

Thirteen letters, thirteen Democrats. Foodborne illness used to be the least partisan subject in Washington. Rosa DeLauro wrote FDA about a Cyclospora outbreak traced to this same part of Mexico in 2013, asking on what legal authority the government could identify a source and decline to name it, and wrote again in 2018 about another one. Nobody at the time thought counting sick people was a party matter. Every tool I have described here belongs to the majority. The chairman’s signature on a PSI subpoena, the Oversight majority adopting Garcia’s request, a hearing date, a vote to strike Section 780. None of it requires a single Democrat.

Thirteen letters in a month. Seven dated deadlines expired. Nine good questions from Senator Warren due back on August 27, which will pass the way the other seven did unless somebody with a gavel decides otherwise. As of last night, the fifty states were reporting 29,912 cases and the two agencies attribute 9,481 of them to this lettuce. Two people are dead. I am watching, and so are twenty-nine thousand nine hundred and twelve other people, and not one of us chose to be part of it.

The fifty state health departments now report 29,798 cases of cyclosporiasis this season, up 1,424 from the 28,374 published yesterday. One state moved, and it moved by more than any state has moved in a single day all summer. Michigan reported 13,909 cases and 314 hospitalizations on Thursday.

That is 1,424 cases in one week, and it arrived a week after Michigan told its residents to resume routine guidance for lettuce and salad greens. Both things can be true, and both are. The state health department says new case reports and emergency department visits are declining, and the weekly pace has in fact eased, from roughly 285 a day to roughly 200. It is still 1,424 people. Michigan records about fifty cases of cyclosporiasis in a normal year. It just booked twenty-eight years of them in seven days, and it is still the largest weekly increase of any state in the country. The largest age group in the Michigan data is thirty to thirty-nine, with 2,615 cases, which is to say working adults who eat lunch somewhere.

One comparison in Thursday’s numbers deserves to be pulled out and looked at. CDC reports 13,895laboratory-confirmed cases of cyclosporiasis for the entire United States since May 1. Michigan, by itself, now reports 13,909. One state has passed the national laboratory-confirmed count. The two figures are not the same measurement, and nobody should pretend they are, because Michigan counts probable cases and CDC does not. That is the point. The gap between what a state can see in its own residents and what the federal surveillance system has managed to confirm is now wider than the federal number itself.

No federal number moved. FDA still reports 6,358 illnesses, 278 hospitalizations and two deaths in fifteen states, stamped August 5, which is now eight days old. Maine issued a public health advisory on August 7 saying it meets the criteria for the multistate outbreak, and six days later the federal advisory still says fifteen states and does not name it. CDC’s national surveillance page was updated August 11 and still reads 13,895 laboratory-confirmed plus at least 10,455 awaiting further investigation, a total of 24,350, covering data through August 10 across 47 states plus the District of Columbia and Puerto Rico.

The states are now 5,448 cases ahead of the combined federal figure, against 4,024 yesterday. Some of that widening is the calendar. CDC publishes once a week, on Tuesdays, so the gap opens from Wednesday through Monday and closes every Tuesday when the federal count catches up, and anyone quoting the gap should say which day of the week they are standing on. The rest of it is Michigan, and that part will not close on Tuesday.

The thirteen states that publish a hospital figure now total 735 between them, after Michigan moved from 279 to 314. CDC’s national surveillance, updated August 11, reports 740 hospitalizations among its laboratory-confirmed cases. Those two numbers are now five apart. A floor assembled by hand from thirteen state health departments has landed within five of what the federal surveillance system counts from a different set of cases, which is either a coincidence or a sign that the bottom-up method works. The outbreak-specific federal figure is a different number again. It is 278, and Michigan by itself reports 314. Ohio is second at 172, and that number has a trajectory of its own: 28 Ohioans hospitalized as of July 2, 96 by July 17, 172 by early August. It is also the oldest figure in this total, because Ohio has not published a newer one, so read 735 as the conservative end. Thirty-seven states publish no hospital figure at all.

Where all fifty stand

The last column of this chart is that state’s entire 2025 total from CDC’s weekly notifiable-disease tables, which is a different series from the state figures beside it. It counts travel-associated cases and covers a full calendar year against a season still running. It is there for scale, not for arithmetic.

What the 2025 column shows

Michigan reported 50 cases in all of last year and has reported 13,909 this year, two hundred and seventy-eight times as many. Ohio reported 76 and is at 4,900, the sixty-four times figure its own health department has used. Indiana reported 27 and is at 1,893. Kansas reported 29 and is at 748. West Virginia reported 8 and is at 296. Arkansas reported 13 and is at 252. Kentucky reported 45 and is at 620.

The same column shows which states are not unusual, and that has to be said. Texas reported 512 cases in 2025 and is at 198 this year. Colorado reported 205 and is at 150. Louisiana reported 99 and has published a single-digit count. Maryland and New Jersey are both running below last year. New York reported 694 and is at 898, which is an increase, but not the kind Michigan and Ohio are having. Not every state on this chart is having an outbreak, and the chart should not be read as though they are.

Four states do not make cyclosporiasis notifiable at all, so no figure exists to compare: Pennsylvania, Idaho, Mississippi and Nevada. Twelve states have published no 2026 count. Two of those, Mississippi and South Carolina, received the recalled lettuce.

What this figure is, and what it is not

29,798 is a floor, not an estimate. It is the sum of what fifty health departments have published, on schedules running from weekly to never, with four states reporting only a range that I count at the bottom. Where a state publishes two series, I use the one this chart has always used: for Florida that is the since-May-1 figure of 342 rather than the calendar-year 352, and for New York the since-May-1 figure of 898 rather than 932. Ohio and Iowa and Kansas and Arkansas are calendar-year, because that is the series those states publish.

Two states publish enough detail to see inside their own number, and both show the same thing. Virginia reports 197 cases and breaks them down: 107 acquired in this country, 48 acquired abroad, 42 unknown or still pending. Roughly a quarter of the Virginia figure has nothing to do with lettuce from central Mexico. Florida is starker. Of its 342 cases, 201 were acquired in Florida, 31 in another state, 45 outside the country, and 65 have no known source at all. Under sixty percent of the Florida number was acquired in Florida.

Some version of that is true in every state on this chart, and forty-eight of them publish nothing that would let anyone check. A floor assembled from fifty different definitions is still worth having, because the alternative is waiting six weeks for a federal number that is itself an undercount. It is not worth mistaking for a case count.

Most people with cyclosporiasis never get a stool test, because the parasite is missed by the routine ova and parasite exam and has to be requested by name, and because people with a diarrheal illness that comes and goes for a month often never see a doctor. The standard estimate of that gap comes from Scallan and colleagues, Foodborne Illness Acquired in the United States, Major Pathogens, in Emerging Infectious Diseases in 2011, which puts the underdiagnosis multiplier for Cyclospora at 83.1. Applied to CDC’s 13,895 laboratory-confirmed cases, that puts the real number near 1,155,000.

What to expect

This chart will cross 30,000 within days. It stands 202 short tonight, and West Virginia posts Friday, Missouri Sunday, Arkansas and Virginia Monday, CDC and North Carolina and Oklahoma Tuesday, Kansas and New York next Wednesday. Ohio is weekly and overdue, last published at 4,900 on August 4. Indiana updates on weekdays and was near 1,900 earlier this week. Any two of those get it there.

Two things will keep it climbing after the illnesses stop. CDC estimates a six-week lag between the day someone gets sick and the day the case reaches it, so August onsets will still be arriving in the middle of September. And the confirmed and unconfirmed pools are still sorting themselves out: on Tuesday the confirmed count rose by 3,427 while the pool awaiting investigation fell by roughly 1,800, which was mostly laboratory confirmation catching up with cases already counted rather than new illness. Expect more of that, and expect it to be reported as a surge.

The season closes on August 31, eighteen days from today. The counting does not.