The retention rule for a seafood importer's verification file is one sentence in one paragraph, and it is easy to get wrong in a way that quietly destroys the records most likely to be asked for. This piece follows the importer's records, and only those, from the section that creates them to the paragraph that says how long they live.
Which records are these?
21 CFR 123.12(c): "The importer shall maintain records, in English, that document the performance and results of the affirmative steps specified in paragraph (a)(2)(ii) of this section. These records shall be subject to the applicable provisions of §123.9."
So the records are the evidence that a step was done and what it showed. For step (a), the lot's HACCP and sanitation monitoring records you obtained. For (b), the certificate. For (c), the visit report. For (d), the plan and the guarantee, and whatever you wrote when you read the plan. For (e), the test results and the guarantee. For (f), whatever the described measure produces.
The written verification procedures and the product specifications are not on that list. They are, in §123.9(c)'s words, "plans and procedures required by this part," which must be "available for official review and copying at reasonable times." No retention period attaches to a procedure. It has to be current. But the record of a step performed in 2025 was performed under the 2025 version of the procedure, and a reader of that record in 2027 will want to see the procedure it was performed under. Keeping superseded procedures with the records they governed is not required by the text and is the only way the records make sense.
How long?
§123.9(b)(1): records "shall be retained at the processing facility or importer's place of business in the United States for at least 1 year after the date they were prepared in the case of refrigerated products and for at least 2 years after the date they were prepared in the case of frozen, preserved, or shelf-stable products."
Two clocks, chosen by the product the record is about. Fresh tuna, one year. Frozen shrimp, two. The regulation does not say "one year," and it does not say "two years"; it says which one for which product, and a file that applies one period to everything is wrong for half the roster.
§123.9(b)(2) adds a third clock for a kind of record importers hold more often than they realise: records "that relate to the general adequacy of equipment or processes," including "the results of scientific studies and evaluations," are kept "for at least 2 years after their applicability to the product being produced." A validation study a processor sent you to support a critical limit in its plan is this kind of record. So, arguably, is your own review of that plan against the Guide, since it is an evaluation of the adequacy of a process. The safe reading is two years after the plan it evaluated stops being the plan in force.
From when?
"After the date they were prepared." Not the shipment date, not the entry date, not the expiry date on a certificate. The clock starts when the record came into existence, which for an importer's verification record is the date you performed the step and wrote it down.
This has one consequence worth stating. A continuing certificate obtained in January 2026 and valid to December 2026 is a record prepared in January 2026. For frozen product it is kept until at least January 2028, a year past its own expiry, because the question an investigator asks is not whether the certificate is current but whether the step was performed when the product was entered. The expired certificate is the evidence that it was.
What about a mixed roster?
Most seafood importers buy fresh and frozen product from the same processor, sometimes the same species in both forms. A plan and guarantee held under step (d) then serves a one-year row and a two-year row at once. There is no rule for this because the regulation is written per product; the practical answer is that the document is kept for the longer period, since it is the record for the frozen product too.
The failure mode here is software. A records system, or a filing habit, that assumes twelve months will delete the frozen-product records first, and frozen product is the majority of what most importers bring in and the majority of what the Guide's aquaculture-drug and pathogen chapters are about. Twelve months is the wrong default. The right default is the product.
Where?
"At the processing facility or importer's place of business in the United States." §123.9(b)(1) again. The importer's records live at the importer's place of business; it is the importer's file, not the processor's, and the regulation does not ask the importer to hold the processor's own records in the US except where step (a) makes them the importer's evidence.
§123.9(b)(3) allows records to be moved "to some other reasonably accessible location" and "immediately returned for official review upon demand." Read the paragraph before relying on it: it is written for a processing facility "closed for a prolonged period between seasonal packs," a processing vessel, or "a remote processing site." An importer's office is none of those. An importer that stores records offsite is doing something the paragraph does not address, and should be able to produce them at its place of business in the ordinary course.
Electronic records change the meaning of "at" without changing the rule. §123.9(f): "The maintenance of records on computers is acceptable, provided that appropriate controls are implemented to ensure the integrity of the electronic data and signatures." Part 123 does not define the controls. The reference point elsewhere in FDA's regulations is 21 CFR Part 11, and the controls it describes are the obvious ones: accounts that identify one person, timestamps the person cannot set, a stored meaning for each signature, and a change history. An importer using a system with those controls can say what "appropriate controls" it implemented; one using a shared spreadsheet cannot.
Copies are fine. FDA's Q&A guidance says records "need not be originals" but must be "accurate and legible" (X.E.7), and that importer verification records "must be maintained in English" (X.E.4).
What happens to the clock after a warning letter?
Nothing, in the regulation. §123.9(b) does not extend retention because of an inspection, a 483 or a letter. But consider FDA's May 26, 2026 letter to a Houston importer, which called its finding "a repeat observation from the previous inspection conducted 08/28/2019 through 08/29/2019." The record that would have answered that letter is the one showing what the firm did in 2019 and after, and by 2026 the regulation had long since allowed it to be discarded. Retention is a floor. An importer that has ever had a §123.12 observation has a reason, not an obligation, to keep the records that show what it did about it for as long as the firm exists.
What the rule does not say
No deadline for producing records; the standard is "reasonable times," and "immediately" applies only to the seasonal-storage case above. No required format. No requirement that records accompany the product at entry. Those three are covered in the companion piece on §123.9, along with the rest of the section.
Citations: 21 CFR 123.9(b), (c), (f); 21 CFR 123.12(c); FDA, Questions and Answers on the HACCP Regulation for Fish and Fishery Products, X.E.4 and X.E.7; 21 CFR Part 11 as a reference point only; FDA warning letter to a Houston importer, May 26, 2026. Nothing here is legal advice. Where this piece and the regulation differ, the regulation is right and we would like to know.