Judge Orders Disclosure of Breath Test Source Code; Thousands of DWI Tests Could be Thrown Out
The Honorable Jack S. Nordby, a District Court Judge in Hennepin County, Minnesota, has ruled that disclosure of the Intoxilyzer source code was “not only relevant to this litigation (and to virtually any dispute involving Intoxilyzer evidence), but very likely indispensible [sic] to a proper hearing of such a case.”
In his order in the case of Minnesota v. Gadow (27-CR-08-46085, a full copy of which is below), the Court states that there is “good reason to believe these source codes will reveal defects in the Intoxilyzer device’s operations of a significance that makes the evidence exculpatory and constitutionally subject to mandated disclosure by the prosecution.” The Court also notes that even though the significant amount of evidence leads to the conclusion that there is a high likelihood that Source Code review would uncover previously hidden errors in the results of DWI tests, a much lower standard should be applied in a discovery motion. “No such showing,” the Court reasoned “need be made to support a simple discovery request as that under consideration here.”
The Court also commented that, when designing the Minnesota Model of the Intoxilyzer 5000, the State chose not to include an option that would preserve a driver’s breath sample. As a result of this choice by the State, a driver’s breath sample is discarded immediately after the machine finishes its secretive analysis, making further testing impossible. According to the Court, the State’s decision to opt out of the preservation feature on the Minnesota Model “results in routine destruction of potentially exculpatory evidence.”
Because of the design of the breathalyzer, the nature of computer construction and operation, and the destruction of a driver’s breath sample, “the potential for errors is enormous,” according to the Court, “the number of possible defects astronomical – literally in the millions or billions” and “[m]any, if not most, if not all of these are virtually immune from detection unless the source codes can be examined.
Continuing in its analysis about the mysteries of CMI’s black box, the Court noted that the courts have delegated their gate-keeping function of determining if the machine is functioning the way it is supposed to be to the machine itself. As a result, police officers testifying that machine is working properly only believe that because the machine said it was working properly. Such a delegation, according to the Court, constitutes an “abdicat[ion of] our non-delegable judicial duties, demeaning the separation of powers.”
The Court also noted that the breath testing landscape in Minnesota presents a situation “where juries and judges are asked to accept the word (virtually the verdict) of a machine that a person is guilty of a crime […] without being allowed to know how the machine reached its decision, or whether it did so reliably.” With the machine not being subject to cross-examination and destroying the breath sample provided by the driver, a serious question must be considered as to whether it results in a denial of a driver’s constitutional rights to confront their accusers and present a full and effective defense.
Concluding its analysis, the Court notes that what was before it was a simple discovery motion, “a simple and reasonable request by the defense to examine the fact-maker and surrogate fact-finder.” The Court held that not only are drivers entitled to Source Code review, but goes on to say that “defense [attorneys] would arguably be guilty of ineffectively assisting a client for not demanding the information.”
A PDF version of Judge Nordby’s order is available here, and the full text is included below.
STATE OF MINNESOTA DISTRICT COURT
COUNTY OF HENNEPIN FOURTH JUDICIAL DISTRICT
State of Minnesota,
Plaintiff,
vs ORDER ON DISCOVERY MOTION
FILE # 27-CR-08-46085
Joshua Thomas Gadow,
Defendant,
Defendant Gadow moves for an order requiring disclosure of the Intoxilyzer
source codes in this criminal impaired driving case.. Aware of the decision in State v.
Underdahl, 749 N .W.2d 117 (Minn . Ct. App . 2008), review granted (Aug: 5, 2008),
where the court emphasized that no showing had been made to support the claim that the
requested material was relevant of would lead to information useful to the defense, Mr.
Gadow has submitted a good deal of material, the pages of which (although I have not
actually counted them) I estimate to number perhaps fifteen hundred to two thousand.
I have examined these. They are for the most part informative as to the questions
raised in Underdahl, and often of great interest. The documents consist, inter alia, of
scholarly articles, affidavits, transcripts of testimony of expert witnesses, and judicial
opinions from several jurisdictions. From their particulars, and from their cumulative
effect, I have no doubt that the source codes are not only relevant to this litigation (and to
virtually any dispute involving Intoxilyzer evidence), but very likely indispensible to a
proper healing of such a case. There is, indeed, good reason to believe these source
codes will reveal defects in the Intoxilyzer device's operations of a significance that
makes the evidence exculpatory and constitutionally subject to mandated disclosure by
the prosecution, though no such showing need be made to support a simple discovery
request such as that under consideration here.
These materials, considering primarily if not exclusively the sworn statements of
experts whose credentials are excellent for the purpose, show (in very brief summary) the
following:
1 - The Intoxilyzer is a complex, sophisticated device designed to measure the
amount of alcohol in air samples. (Head and Workman, "Litigation in the United States",
pages unnumbered)
2 - This device is "computer controlled;" it is, in fact, effectively a computer
itself, or a complex of computers. (Ibid).
3 - The device, therefore, (and each "computer" it comprises), is controlled by
“software,” including the "source codes." (Ibid).
4 - These codes are the instructions that dictate to the device how it must operate .
(Ibid)
5 - The machine itself has no autonomy, no discretion. It does what the software
tells it. (Ibid).
6 - Therefore all similar machines governed by the same codes will operate
identically (assuming no defects or variations unique to any of the machines are present),
(Ibid).
7 - Virtually all software has defects . (Ibid).
8 - Such defects can cause malfunctions in the Intoxilyzer’s operations, and
instances of those have been proved. (Ibid).
9 - Software will cause similar machines to reach identical results A software
defect, accordingly, can cause identical erroneous results in multiple machines, That is,
similar results from several machines are, in themselves, no proof of accuracy except
where it is known the software is defect-free. (Ibid).
10- “A computer scientist… can frequently find defects in software by ‘reading’
the source code.” (Ibid). Codes may also be reviewed "automatically," Either method
can identify flaws in the source code which render Intoxilyzer results inaccurate. (Ibid).
11 - The process of translating "source code" to "machine code", an essential part
of the Intoxilyzer operation, involves numerous possibilities of error, in effect a result of
ambiguities in instructions which can lead to contradictory or erroneous results. (Ibid).
12 - Faults inherent in and resulting from defective software or source codes may
affect various aspects of the Intoxilyzer's operation, including: A) determining whether a
test was actually attempted or conducted, or "refused," B) determining whether an
"adequate' sample was given, C) determining whether the proper portions of the sample
("alveolar" air) are tested, and other portions excluded, D) determining that "interferents"
(hydrocarbons) are properly excluded from the analyzing process (a failure to do which
can result in improper elevation of a reading), E) determining whether a simulation is
properly analyzed. These potential defects might lead to faIse and questionable readings,
either as to quantity of alcohol or as to whether an inadequate sample was given (and thus
a “refusal” committed).
13 - The machine's ostensibly successful analysis of known samples or
simulation gases does not demonstrate its reliability in analyzing unknown subject
samples, for the latter may and often do contain other substances, not present in the
controlled samples, which must be accounted for and eliminated from consideration in
the subject tests. Since the samples are destroyed in testing there is no way to determine
whether this function is properly performed, or if the device is capable of doing so
without the codes.
It is notable also that the Intoxilyzer discards samples once analyzed; they are not
retained to allow additional testing, This is by choice of the manufacturer and Intoxilyzer
purchasers. It results in routine destruction of potentially exculpatory evidence. "Breath
testing stands alone in the forensic sciences, as the only forensic testing method that
prohibits verification and validation by virtue of the design of the process." (Ibid).
Moreover, no records are ordinarily kept by manufacturers or law enforcement agencies
of defective tests; and there are no reporting mechanisms for failures. (Ibid).
The nature of computer construction and operation, both software and hardware,
is so complex (and rapidly changing) that the potential for errors is enormous, the number
of possible defects astronomical - literally in the millions 01 billions.
Many, if not most, if not all of these are virtually immune from detection unless
the source codes can be examined.
One expert with sterling credentials has documented defects in an Intoxilyzer and
said that "this obviously serious problem in data integrity makes it essential that the
sources of the software be thoroughly examined to identify the source of these serious
errors and their potential effect on all data." (T, Burr. Affidavit, December 17, 2007). In
another affidavit the same expert states that access to source codes is necessary to
evaluate the functioning of Intoxilyzer, and "without access to those codes it is not
possible to determine if the Intoxilyzer functions as designed or as approved by rule "
(Burr affidavit, October 11, 2007).
Another well-qualified expert states that because the manufacturer has not made
the source codes available, the device's "process" of breath analysis is not fully known,
and "the exact process can only be determined by examining the source code of the
software program running in the instrument." (H. Myler affidavit, September 9, 2006),
Evidence was also presented that a different device for analyzing breath samples
operated with a source code defect affecting the "slope detector" that caused numerous
failures, (Simpson et al. article; Kierzek testimony transcript).
Moreover, e-mail correspondence between the Minnesota Bureau of Criminal
Apprehension and CMI, the manufacturer, reveals that the Intoxilyzer’s determination of
"deficient" breath samples varies according to the software being used, and it is unclear
whether this problem has been satisfactorily corrected
Under present Minnesota law, a substantial (even alarming) portion of the truth –finding process in both criminal driving and civil licensing cases, otherwise normally
performed by judges and juries, is effectively delegated to the Intoxilyzer machine,
which is in turn a tool selected, obtained, maintained, and operated by the executive
blanch of government, more specifically its law enforcement branches, with the blessing
(so to speak) of the legislature. Virtually no one who is directly involved in the courtroom
aspects of these cases (police officers, prosecutors, defense lawyers, defendants, judges) has the scientific knowledge or training to understand how this machine works.
We are asked, in other words, to have faith in the manufacturer's and the executive
branch's decisions and judgment in certain important respects, to assume that if a
minimally trained police officer follows a number of relatively simple procedures in
operating the machine, it will yield a reliable result both as to sufficiency of samples and
alcohol readings To the machine itself we largely delegate the gate-keeping function of
deciding whether the machine is functioning correctly. The officer who testifies about
the test (who is often the only witness) has no way to determine whether it is working
properly, except from what the machine tells him. (We thus abdicate OUI non-delegable
judicial duties, demeaning the separation of powers.)
But what the machine tells him (and tells a fact-finder) is what the programmers
designed it to say. Even setting aside any suspicion of deliberate deception, assuming no
bad faith is involved, this programming (which for present purposes we may assume
consists in significant part of the "source codes" in question) could lead to undetectable
misrepresentations that are virtually conclusive of the result in litigation.
For example, (to state only one), in a state where the decisive blood-alcohol
reading is .08, any quantity below that would be exculpatory. The controlling software
for the device could be designed to instruct the machine to round any reading slightly
below that (say .07999) to .08.. Since this instruction would be only in the code, and in no
way detectable from either examination of the machine itself, questioning of the operator,
or other reasonably available means, it would be shielded from discovery, disclosure, or
diagnosis.. Innocent people would be convicted, with no recourse,
We have therefore a situation where juries and judges are asked to accept the
word (virtually the verdict) of a machine that a person is guilty of a crime (of either
driving with more than .08 blood alcohol or of refusing a test), and we are asked to accept
these determinations without being allowed to know how the machine reached its
decision, or whether it did so reliably. Although it is not the direct and immediate
question here, this raises a serious concern that the accused is denied the right to confront
his accuser (the machine), and the due process right to present a defense. The machine's
hearsay statement is not subject to cross-examination. The machine itself destroys (or
does not preserve) the sample, making further testing impossible
But this proceeding is a mere discovery motion, a simple and reasonable request
by the defense to examine the fact-maker and surrogate fact-finder. Not only is the
defense entitled to this information, defense counsel would arguably be guilty of
ineffectively assisting a client for not demanding the information.
No even marginally persuasive or even plausible reason to deny this threshold
investigative request has come to my attention, either in the evidence or arguments in this
case, or in the opinions of the Minnesota and other courts that I have examined.
It is worth repeating that this is a mere, reasonable discovery request, a routine
pre-trial step in preparation for litigation or negotiation, a perfectly rational inquiry into
the foundation for the result of a scientific test. The refusal of the manufacturer, and the
refusal or inability of the prosecution to comply with these requests, and ensuing court
orders, have caused totally unjustified disruptions in a justice system that is already
overburdened by lack of adequate funding and resources. The making, opposing,
hearing, deciding, and appeals of these motions have consumed untold amounts of time
of all involved, have caused a great deal of expense, have prompted certain unwise
judicial policies designed to deal with the volume of these motions, have resulted in a
plethora of often conflicting, often contradictory decisions in both trial and appellate
courts, have prevented the resolution of cases within reasonable times, and have in
general as a consequence diluted the quality of justice and beschmirched the appearance
and reputation of the machinery of justice. All of this is direct result of a private, profit-making corporation's desires and efforts to withhold from courts pertinent information
about the product by which it makes its profits, a product which is incidentally capable
under the current state of the law of convicting people (in very large numbers) of crimes.
That this is true - and not the plot of a futuristic novel or screenplay - is astonishing. It is
also shameful, since it is for the most part easily avoidable, by proper orders in individual
cases requiring disclosure on pain of suppression or dismissal. Instead, however, the
courts have too often been willing to allow an interested non-party to foul the gears of
justice, and leave countless otherwise resolvable cases in a limbo of uncertainty, creating
an unnecessary back-log of cases that will continue voraciously to consume resources
The defense motion is granted.
The source codes shall be produced to the defendant within ten days of the date of
this order or the test results shall be suppressed .
It shall be no excuse for non-compliance that the prosecution does not have the
source codes. This issue has long been familiar. The prosecution is the proponent of the
evidence, and cannot frustrate the defense's right to the evidence (and this court's order)
either by failing to obtain the source codes, or by agreeing or acquiescing with the
manufacturer's unjustified refusal to produce them. This case may proceed the old- fashioned way, by presentation of admissible evidence to human fact-finders.
IT IS SO ORDERED.
Dated: November 18, 2008
Will there be any television news coverage of this on local MN news stations?