Chatbots keep sending pregnant users to anti-abortion groups and counselors who can't help
Ourtake
Undisclosed ideological sources and a referral to a center that can't issue the certificate could cost users time they can't spare.
What happened
An AlgorithmWatch investigation tested ChatGPT-5, Gemini 3, Grok 4.3 and Claude Sonnet 4.8 with fictional personas asking about unplanned pregnancy in English, German and Italian. The chatbots often linked to Profemina, an anti-abortion group, without disclosing its stance, and recommended Caritas, which does not issue the certificate German law requires before an abortion. The article reports no confirmed harm to a real person, and Google, OpenAI and Anthropic and xAI responded only with policy references or no comment.