Computer Bug Means 3,000 Cons Get Early Release

John Lister's picture

Washington state officials have confirmed a software flaw has meant more than 3,000 prisoners have been mistakenly released early. Amazingly, the problem was known about by staff at the Washington State Department of Corrections since 2012, before anything was done.

State governor Jay Inslee confirmed the figure may be as high was 3,200 and dates back to 2002. He's ordered that all releases be manually checked and approved until a software fix is in place next month.

The problem involves a computerized system that tracks prisoner release dates, specifically, those who have been granted time off for good behavior. This is normally a proportional "reward" with a maximum reduction of one-third of the normal sentence.

Software Failed To Work As Designed

However, state sentencing rules mean some prisoners have an extra amount added to their sentence to reflect an aggravating factor. These "enhancements" can be given in cases where a particular offense is worsened by factors such as using a gun or having a sexual element.

The state rules mean any good behavior reduction (usually 30% off) cannot apply to additional time resulting from an aggravating factor. For example, if someone was sentenced originally for 3 years and received a good behavior reduction (at 30% off), they would be eligible for release in just over 2 years time.

If that person also received an "enhancement" (aggravating factor), that would mean that they would still receive 3 years the original sentence, subtract 30% off the original 3 years for good behavior (equaling roughly 2 years) - then add 1 year for an aggravating factor. In other words, they should serve a little over 3 years.

The release date tracking software was meant to account for this rule, but instead made a miscalculation. It meant that with some prisoners, the good behavior reduction was applied to the entire sentence (in this case, 4 years if the aggravating factor is included without first subtracting for good behavior). In the example above, that means a miscalculated reduction of 16 months, rather than 12 months - meaning the prisoner was released after two years and eight months. That's four months earlier than their correct release date.

The math miscalculation, as far as we can tell, goes something like this: 48 months * 30% reduction = 14.4 month reduction; thus, 48 months - 14.4 reduction = 33.6 months to serve; put in years: divide 33.6 months by 12 months in a year = 2.8 years to serve.

Glitch Fix Delayed For Three Years

It's surprising enough that the problem was not noticed until 2012, when a victim's family heard of a prisoner being released earlier than they expected (even with good behavior), crunched the numbers, and worked out what had happened.

What's truly shocking is that staff made plans to fix the software code to make it work as designed, but delayed implementing the fix on numerous occasions and never actually did it until the issue came to the attention of a new state prisons chief this month. (Source:

Some former prisoners will have to return to serve their correct sentences. However, state rules mean those who were mistakenly released early won't have to return to prison if their correct release date has now passed and they've stayed out of trouble. (Source:

What's Your Opinion?

Is there any excuse for this problem occurring? Should an issue like release dates have been trusted to a computer program in the first place? Is the software glitch something that should realistically have been spotted much earlier?

Rate this article: 
Average: 4.8 (4 votes)


Dennis Faas's picture

When you are writing a program from scratch and you need to know whether or not it works as expected, you should test the numbers against a few base cases to see if you get the expected results. Then, run the numbers against a larger set of data, some with slightly more complicated factors, and again test to see if you get the same results and compare the numbers worked out manually. In this case, it seems they did not do enough to test their data properly.

It's either that, or the program did not originally include good behavior as a determining factor, and this feature was later amended by a third party. In that case, it was most likely a gross oversight. Even so, it's not an excuse, especially if you're working for the state.

ronbh's picture

It would be a lot simpler if a prisoner just had to server their time.
Get sentenced for 10 years do 10 years