Congress's own research service published an AI companion primer

The design

On 14 August 2026, the Congressional Research Service (CRS), the nonpartisan analytical arm of the U.S. Congress, published Report R49189, AI Chatbots as Companions: Overview, Uses, and Considerations for Congress. The report is available through CRS's own report page at Congress.gov and, in mirrored full text, through EveryCRSReport.com's archive of CRS products. It defines a chatbot as software that simulates human conversation, traces the category back roughly 60 years to ELIZA, and separates 'companion chatbots' built for entertainment, emotional support and eldercare from general-purpose chatbots that users repurpose for the same ends.

What the evidence says

The report is explicit about the limits of current knowledge, stating in its own summary that 'much of what is known about these AI chatbot interactions is from short-term studies or company disclosures.' That is CRS's own assessment, not this site's characterization, and it is the report's central evidentiary caveat: it presents no new data of its own, and it notes that some user-reported experiences are positive while separately citing published accounts describing psychological distress and reported deaths associated with chatbot use, attributing each claim to the outlet or study that made it rather than asserting either as an established CRS finding.

What it asks of people

As a report addressed to lawmakers, CRS asks Congress, not individual users, to weigh legislative options it lists: requiring disclosure of a chatbot's nonhuman status, age verification for companion products, design features that prompt users to take breaks, and limits on processing user data for chatbot outputs. It recommends Congress consider funding longitudinal, interdisciplinary research, through agencies such as the National Science Foundation or National Institutes of Health, running longer than a year, specifically because the existing evidence base was not built for that timescale.

Privacy and safeguards

The report catalogs proposed rather than enacted safeguards: bills introduced in the 119th Congress that would require crisis-resource links and nonhuman disclosure labels, restrict minors' access to AI companions, and limit data processing, none of which CRS presents as current law. It consistently distinguishes what 'some Members have introduced' from what has passed, keeping its account of the legislative landscape separate from its account of the underlying research gap.

• Which of the bills CRS describes have since become law, and which remain proposals?
• Does a claim about chatbot harm cited elsewhere trace back to one of the short-term studies CRS says the field currently relies on?
• What would a longitudinal study funded along the lines CRS recommends need to measure that a short-term study cannot?

CRS's own framing is candid about an evidentiary gap: a nonpartisan research service telling Congress that the research base for AI companion chatbots is thin is itself a data point, distinct from any single study's finding, about how early this field of evidence still is.