We’ve wondered if the widespread use of generative AI was going to produce an increase in the filing of pro se complaints and a marked change in their content. While the results are preliminary, at least one study shows a significant increase in pro se filings following the general availability of AI tools. And our own anecdotal experience suggests that, with the benefit of AI, pro se filings are moving from handwritten one-pagers to lengthy and more sophisticated complaints.
If AI is helping pro se plaintiffs prepare more sophisticated pleadings, we would hope that courts would become less deferential. After all, if generative AI helps pro se litigants bring their claims on a more level playing field, and if they’re bombarding courts with complaints, shouldn’t we see an end to the type of judicial indulgence in favor of pro se plaintiffs that leads to bizarrely pro plaintiff results? While we will continue to hold out hope, today’s case is an example of a court giving a pro se plaintiff the benefit of the doubt and, unfortunately, potentially expanding Texas product liability law into enterprise software.
Today’s decision involves claims against the manufacturer of an electronic health record system utilized by hospitals. Fischman v. Epic Systems Corp., 2026 WL 1651510 (N.D. Tex. June 8, 2026). The pro se plaintiff claimed that the defendant’s electronic record system did not display accurate medical information necessary to treat her father’s medical condition, and that he died as a result. While we don’t know for certain that the complaint was generated or edited by AI, it’s a detailed, 25-page document with 109 numbered paragraphs, including claims for wrongful death, negligence, design defect, negligent misrepresentation, and gross negligence. Certainly not your run of the mill pro se filing. When we asked our AI tool to review the complaint, the tool predicted there was a strong chance that AI generated it. And the complaint certainly reads like it was drafted by AI. These are important points when considering whether the pro se pleading deserved any extra deference.
Defendant moved to dismiss, arguing that it did not owe the plaintiff any duty and that its electronic records platform was not a “product” under Texas law. The defendant argued that no Texas court and no court within the Fifth Circuit ever held that enterprise software constitutes a “product” subject to product liability claims. The Texas product liability act does not define “product,” but Texas courts look to the Restatement for guidance on that issue. The Restatement defines product as “tangible personal property distributed commercially for use or consumption.” Restatement (Third) of Torts: Products Liability § 19(a) (1998). That definition would not apply to electronic data relied on by physicians and nurses in the care and treatment of patients. While the Restatement includes a narrow exception for “intangible” products, the Texas Supreme Court previously rejected “intangibles” as products under the Restatement. See Firestone Steel Products Co. v. Barajas, 927 S.W.2d 608, 616 (Tex. 1996) (discussed in our 50-state survey on the treatment of electronic data as a “product”).
But rather than citing to the Texas decisions that rejected the application of the “intangible” exception in the Restatement, the decision relies on a footnote from a Texas appellate court decision that assumed software could be a product for the purposes of that case because it was a “highly technical tool,” and then dismissed the case based on the economic loss rule. Fischman, 2026 WL 1651510 at *5. Then, with little discussion of the limits federal courts normally apply when making Erie predictions that expand state tort law (see our post on how the Erie doctrine is supposed to work), and based in part on favorable “inferences” drawn in favor of the pro se plaintiff, the court predicted that the Texas Supreme Court would treat defendant’s electronic health record system as a “product.” Given that prediction, the court declined to dismiss the products liability claim. Id. at *6.
For the alternative feasible design prong of the products claim, the decision assumed that a hypothetical software program without the alleged deficiencies would constitute an acceptable alternative design. That treatment of alternative design could be said for any complex software system, which is one of the reasons such programs are not analogous to or treated as products for purposes of product liability claims. It is also inconsistent with the general treatment of alternative feasible design in product claims (see our primer on alternative feasible design here, and other alternative feasible design summaries here and here).
The court also declined to dismiss plaintiff’s negligence claim. Although the complaint did not allege that the defendant acted as a contractor, the court borrowed from contractor-liability principles since the plaintiff alleged that the defendant had some continuing control over the software. Again, that was based on “[c]onstruing the pro se amended complaint liberally.” Id. at *3. That potentially results in a duty on medical software manufacturers irrespective of intervening use by healthcare professionals. It also has the potential to turn run-of-the-mill medical malpractice claims involving the use of electronic medical record systems into product liability claims against medical software manufacturers and sellers.
We hate to leave anyone heading into the weekend on such a sour note. Perhaps this from SRV will help. Texas blues indeed.