On Tuesday I argued that the model act needs a retention clock: a deadline by which the government must delete or return the non-responsive remainder of a forensic image, unless a court says keep it. It’s easy to wave that off as academic. So it’s worth remembering that a federal appeals court once tried to impose exactly that clock — not on one case, but on every digital warrant in the largest circuit in the country — and then, a year later, took it back. Why it took it back is the strongest argument I know that this rule has to come from a statute.
The case is United States v. Comprehensive Drug Testing, and it grew out of the BALCO steroids scandal. The government obtained a warrant for the drug-testing records of ten baseball players it had probable cause to investigate. When it executed that warrant against the testing lab’s computers, it seized and reviewed the records of hundreds of players — and many non-players — none of whom the warrant named. The intermingling of a few responsive files with a mass of non-responsive ones became the pretext for taking, and searching, all of it. If that sounds like every theme of this blog compressed into one seizure, that’s because it is.
In 2009, sitting en banc, the Ninth Circuit did something remarkable. Chief Judge Kozinski’s opinion didn’t just rule that the government had overreached; it announced protocols that magistrates should thereafter require in every warrant for electronically stored information. Read them against the model act and you’ll recognize the architecture, because it’s nearly identical. The government had to waive reliance on the plain-view doctrine, so it couldn’t turn an authorized search into a general one by “plainly viewing” everything its tools traversed. Segregation of responsive from non-responsive data had to be done by specialized personnel walled off from the investigators, or by an independent third party — a filter team. The search protocol had to be designed to uncover only what the government had probable cause to find. And — the piece that matters most this week — the government had to destroy or return the non-responsive data within a reasonable time, and keep the issuing magistrate informed of what it had returned or destroyed and what it had kept.
Stop on that last one, because it is the retention clock, built by a court. Segregate the responsive from the non-responsive; get rid of the rest on a timetable; account to a judge for what you keep. That is § 4 through § 6 of the model act, handed down as binding law for an entire circuit by one of the most respected judges of his generation. For about a year, the experiment was real.
Then it collapsed. In 2010 the Ninth Circuit reheard the case and issued a revised en banc opinion that reached the same result — the government had overreached — but on the narrow, decades-old ground of United States v. Tamura, its pre-digital rule about intermingled documents. The mandatory protocols were stripped out of the binding holding and demoted to non-binding suggestions in a concurrence. What had been a rule the government must follow became guidance the government might consider. The clock, as law, was gone.
Why did it fall? Three forces, and each is instructive. Prosecutors pushed back hard, arguing in their rehearing briefs that under the 2009 standards they were effectively being denied new warrants to search computers at all — the protocols, they said, made digital investigation unworkable. The forensic community argued that the “destroy or return within a reasonable time” requirement misunderstood how digital analysis actually works: examination is iterative, a finding in one file tells you what to look for in another, and a short deadline to purge the non-responsive data can force the government to destroy the very material a later thread would have made responsive. And underneath both ran a deeper legal objection, pressed by scholars like Orin Kerr: that a magistrate issuing a warrant simply lacks the authority to impose these kinds of ex ante procedural rules on how the executive conducts its searches. That is not the judicial role, the argument goes. It’s a legislative one.
Here’s the thing about that last objection: it might be right. And if it’s right, it doesn’t defeat the retention clock — it relocates it. Because if a magistrate can’t impose a return-and-destroy deadline as a condition of a warrant, and an appeals court can’t make one stick as a circuit-wide protocol, then the clock can only come from the one body that unquestionably has the authority to set the rules of criminal procedure: Congress. CDT didn’t fail because the retention clock was a bad idea. A court full of serious judges thought it good enough to impose on an entire circuit. It failed because a court tried to build it, and a court is the wrong builder — too easily reversed under pressure, and quite possibly acting outside its lane to begin with. The model act is the same rule, enacted by the branch that can enact it.
Let me give the objections their full due, because they aren’t throwaways. The forensic point is real: you genuinely cannot always know what’s responsive until you’ve followed the data where it leads, and a deadline set too short, or too rigid, would destroy evidence and free the guilty. The prosecutors’ complaint was real too: a protocol imposed by judicial fiat, without machinery to calibrate it, landed as an all-or-nothing burden that gummed up legitimate investigations.
But look at what those objections argue for. They argue for a well-designed clock, not for no clock. The forensic-iteration problem is answered by a retention period that’s reasonable rather than punitive, extendable on a showing to a judge, and paired with the sealed reference copy § 6 already builds for exactly the case where a later thread might make old data matter — you don’t destroy the possibility of following the thread, you preserve a sealed copy and require a warrant to pull it. The “investigations became unworkable” problem is what happens when a rule is imposed by protocol instead of legislated with its exceptions and calibrations attached. In other words, the reasons CDT’s clock failed in practice are not reasons to abandon the clock. They are a specification for how to draft it — which is precisely the work a statute does and a warrant condition can’t.
That’s why CDT is the most important case this blog has covered that almost nobody outside the field remembers. It is the retention clock’s dress rehearsal: the right idea, tried by the wrong branch, in the wrong instrument, and abandoned within a year — but not before proving that serious judges believe the clock is sound and that the objections to it are answerable by design rather than fatal in principle. The Ninth Circuit showed us the rule, and showed us why a court can’t be the one to write it. Section 6 takes the rule the court sketched and hands it to the branch that can make it last.
Next Tuesday, § 7: use and disclosure — what the government may do with what its search turns up, and the second-warrant problem that has run underneath this whole project.
Leave a comment