The NFA regulates certain firearms (fully-automatic ones, rifles and shotguns with barrels or overall lengths below certain limits, and "any other weapons" (firearms that are neither handguns nor rifles nor shotguns). Each retail transfer of these is taxed, and since the return requires identity, fingerprints, and photograph, the return essentially registers the firearm.
Problems With the Registration Database
A major NFA problem is the poor condition of the database, known as the NFRTR. For the first fifty years of the NFRTR, it was composed of hundreds of thousands of paper files stored in cabinets, with plenty of opportunity for loss, misfiling, or other errors. Over the past few decades it has been computerized, which again poses problems with keypunch errors. The problems with the database first came to light in 1979 Congressional hearings, documented by internal studies that spoke of fears that an innocent man could be convicted due to the errors, and testimony by a licensed dealer that dozens of his guns had been been gathered up for seizure as unlicensed, before he produced his own copies of the registrations. The problems continue to this day.
The errors were highlighted when an internal BATF training film came to light, in which the head of the NFA Branch jested about the errors, acknowledged that the significant error rate had been 50% before he came on, and stated that in spite of them his staff would continue to certify that their records were 100% correct. (The agency thereafter refused to release the tape under FOIA, on the ground that it would invade the official's privacy!) Here's more background on that training tape.
Major internal audits of the NFA database took place in May 1998 and December 1998. For a summary of errors found, click here. For a critique of the audits, click here. (All are pdf scans that load slowly).
In a 2008 prosecution, the defense introduced expert testimony from Dr. Fritz J. Scheuren, an expert on statistics and record keeping. He testified, based on his review of internal BATF audits, that the NFA database was not reliable for purpose of establishing guilt. Click here for the transcript, and go to pages 1016 to 1042.
That trial ended with a hung jury; the defendant then plead to a misdemeanor, and received probation and a $25 fine.
Lack of Testing Standards
This has been another problem area. Agency testimony that the firearm in question fired full automatic is often critical to the prosecution's case, yet the agency has never established protocols for such a test. The persons conducting the test are thus free to keep experimenting until they find circumstances under which a semiautomatic will go full auto -- which can include experimenting with different ammunition, changing lubrication, and other variations -- and then are able to testify that it is a full automatic. For a Congressional Research Service paper on the issue, click here.
Statute of Limitations
As part of the IRS Code, NFA carries a short, three-year limitation. See 26 U.S.C. ¬ß6531. Note that NFA has a possessory element, making it a continuing violation, but if possession cannot be proven within the last three years, the prosecution is time barred.
Other NFA legal issues
Handling transfer of an NFA firearm by inheritance.