Judge Megan Goldish presides over the Domestic Violence Court Division of the Circuit Court of Cook County in Illinois. In February,2026, she had the article “Navigating AI in Domestic Violence Court” published in the CBA Record (a Chicago Bar Association bulletin). As Judge Megan Goldish notes, domestic violence cases, with their issues of trauma and family and power dynamics, represent some of the most complex matters handled within the justice system.
There are a number of uses of AI for litigants and attorneys, including the organization of digital evidence, protective order drafting, and highlighting communications that imply coercive control or stalking. AI also provides a way for self-represented litigants to efficiently translate materials, complete documents, and automate court reminders. It effectively expands people’s access to the justice system.
However, there are also numerous potential risks when it comes to using AI. For one, domestic violence incidents are often underreported, which means that AI models trained using official data may undervalue the actual danger and likelihood of a domestic violence incident occurring. In addition, AI may not be able to make basic distinctions, such as between an abuser and a criminalized survivor. It may also fail in evaluating the complexities involved in various manifestations of domestic violence, including isolation, financial control, stalking, immigration concerns, custody, and self-defense.
Another possible issue is the uploading of confidential data onto unsecured platforms, which potentially endangers the lives of survivors. In addition, pretrial assessments are not an accurate guidepost of fair outcomes in cases, and AI can unintentionally influence pretrial negotiations. For these and other reasons, AI Policy at the Illinois Supreme Court level provides ethical rules applicable to domestic violence situations.
According to Rule 1.4, attorneys must always inform clients on the way AI is being deployed, and on the costs and risks that arise as a result. Rule 1.5 requires that fees are “reasonable and transparent.” This is extremely important, considering the financial vulnerability of many domestic violence victims. Rule 1.6 prohibits uploading confidential client data onto unsecured AI platforms. This rule takes on greater importance when domestic violence survivors are housed in shelters. Inadvertently disclosing that location to the wrong party could endanger their lives.
One of the most prevalent current uses of AI is in pretrial practice, which employs the Public Safety Assessment (PSA) and provides judges with key information when defining the conditions of detention, supervision, or release. Among the elements the PSA weighs are charges, age, prior failures to appear, and criminal history. “New violent activity” is a particularly noteworthy risk flag.
Judges in domestic violence courts look at PSA data alongside a domestic violence screening instrument, which examines other elements of risk, including weapon use, presence of children, substance abuse, prior abuse, and past orders of protection. The screening instrument generates a standardized risk score, which influences pretrial conditions. Uses of AI include distinguishing stalking from accidental proximity when GPS monitoring is in use, as well as analyzing materials such as police reports for signs of abuse.
While such pattern evaluation and recognition has positive applications, it is prone to algorithmic bias, breaches of privacy, and over-reliance on AI when making decisions. It’s critical that judges use human faculties and capacities to ensure due process, accuracy, and privacy, and to avoid false alerts. The author’s takeaway is that no algorithm can ultimately take the place of empathetic, trauma-informed judgement. AI should be used as a support tool, but must never take the place of judicial discretion.
