Pro-Life Groups Fear Push to Regulate AI Chatbots Could Restrict Abortion Alternatives

Efforts to get the European Union to regulate artificial intelligence chatbots will suppress information about alternatives to abortion, pro-life organizations have said.

The warning follows a move by 68 members of the European Parliament (MEPs) to question whether chatbots that direct women toward pro-life websites need regulating and could pose a “systemic risk” under the EU’s AI Act and Digital Services Act (DSA).

An EU Commission spokeswoman explained its regulatory powers in an email to The Epoch Times but did not confirm any intention to use them against pro-life information, nor did it explicitly rule out regulatory action concerning chatbot recommendations.

She said the obligations under the DSA and AI Act “operate in a complementary manner,” with each law addressing different objectives, and that the DSA requires AIs to “assess and mitigate systemic risks” relating to public health, fundamental rights, and other areas.

Under the AI Act, providers must ensure AI does not “engage in any of the prohibited practices” and that it provides sufficient transparency when it “manipulates or generates synthetic content,” she added.

‘Blur the Line’

In tests carried out by The Epoch Times, ChatGPT, xAI’s Grok, Anthropic’s Claude, and Google’s Gemini presented three options when asked about an unwanted pregnancy: keeping the baby, adoption, or ending the pregnancy through abortion.

An investigation by AlgorithmWatch claimed that AI chatbots sometimes “blur the line between official health information and anti-abortion advocacy” when responding to questions about unwanted pregnancies.

The report, “How Popular AI Chatbots Recommend Pro-life Websites to Pregnant People,” tested ChatGPT, Gemini, Grok, and Claude using fictional scenarios in Italy, Germany, and the UK involving women aged 16, 28, and 38 who became “pregnant by accident.”

Across 270 responses, the researchers found that at least one in four queries produced links to what it labeled “anti-abortionist” websites.

The report did not elaborate on what constituted an “anti-abortionist” website or clarify whether these websites included those offering adoptions or just providing pregnancy support without abortion.

The report prompted three written questions that 68 MEPs submitted to the European Commission last month.

The first was, “Does the Commission agree that reliable healthcare information on abortion from official sources is essential to protect the health, safety and fundamental rights of people in Europe and, if so, does it agree that the general-purpose AI chatbots could pose a systemic risk under the AI Act and, in the case of ChatGPT, under the Digital Services Act?”

The second was, “Is the Commission aware of the AlgorithmWatch report? If so, what steps has it taken to investigate and address this issue?”

The third was, “What other steps is the Commission taking to address the reliability of information given by chatbots and even possible foreign influence via chatbots?”

Left-wing and center-left political groups across Europe approved the questions, alongside members of the centrist Renew Europe group and one member of the center-right European People’s Party.

However, pro-life organizations said that attempts to restrict such recommendations could prevent women from finding pregnancy-support centers.

‘Really Want to Hide It’

Andrea Trudden, a representative of U.S.-based pregnancy-support organization Heartbeat International, which was mentioned in the report, told The Epoch Times that she believed this is an attempt to limit the visibility of pro-life organizations online.

The group says it supports more than 4,000 pregnancy-help organizations across more than 100 countries, providing training, education, and practical assistance.

“They are simply not happy that pregnancy centers are coming up as answers to women,” she said.

Trudden said that they “really want to hide it so that women don’t see it as an option.”

AlgorithmWatch claimed the systems frequently failed to distinguish between official healthcare authorities and advocacy groups.

One organization, Profemina, an affiliate of Heartbeat International, appeared in about 17 percent of responses across languages and platforms. AlgorithmWatch said that “the chatbots did not warn the users about ProFemina’s anti-abortion positioning.”

On its site, Profemina says it has counseled more than 800,000 women and that it has learned that “external pressure from circumstances and people is the leading cause for abortion decisions.”

“Our goal at Profemina is to stand with pregnant women in need, devising workable alternatives to abortion with all women who so desire,” it says.

Simon Ostermann of the German Research Center for Artificial Intelligence (DFKI) was quoted in the AlgorithmWatch report saying that organizations like ProFemina have “little risk of being censored by the models.”

“From my view, actors like ProFemina present themselves in respectable, civic-minded language—protecting the family, careful not to be seen marching with far-right parties. That kind of website has no obvious red flags. No racist content. So it doesn’t get filtered out,” Ostermann said.

AlgorithmWatch also claimed that chatbots recommended Germany’s Catholic counseling organization Caritas, which has 300 counseling centers, without explaining that it cannot issue a Beratungsschein certificate, a certificate confirming that a woman has completed mandatory pregnancy counseling in Germany, which must take place at least three days before an abortion.

The Epoch Times has contacted Caritas for comment.

AlgorithmWatch is funded by Pierre Omidyar’s Luminate, Stiftung Mercator, Schöpflin Stiftung, Robert Bosch Stiftung, and the Alfred Landecker Foundation, alongside German government funding programs.

AlgorithmWatch did not call for banning pro-life websites from chatbot results; instead, it called for more transparency and accountability in the results.

It claimed the models frequently failed “to distinguish between official health authorities and advocacy organizations, making it difficult for users to assess the provenance and reliability of the information.”

Trudden said pregnancy-support organizations generally make clear that they do not provide or refer women for abortions.

She said the report reflected “a pattern of behavior that is very, very pro-abortion and very, very anti any life affirming alternatives.”

“We provide information in regard to pregnancy health. We do not refer for nor provide abortion,” Trudden said.

She said that directing women to pregnancy-help resources showed the organizations were answering questions being asked online.

“It’s scouring the internet because this is information that is out there. These are organizations that are there to help. They’re providing education, and so that’s why the chatbots are picking them up.”

Censorship

Adina Portaru, director for Europe at Alliance Defending Freedom (ADF) International, told The Epoch Times by email that the prospect of regulatory penalties could nevertheless encourage AI companies to restrict certain sources.

“This latest attempt at censorship is alarming. When women turn to AI chatbots for information about their pregnancies and the resources available to them, the answers they receive should not be skewed to only show pro-abortion content,” Portaru said.

“Censoring pro-life resources from ChatGPT, Grok, and similar chatbots is concerning, as it keeps women from discovering care options that could truly help them, such as pregnancy resource centers, pro-life medical professionals, and organizations that provide financial assistance to expecting mothers.”

She said that decisions taken by global technology companies in response to European regulation could also affect users outside the EU.

“Because these platforms are global, the effects reach users well beyond Europe. Faced with the threat of heavy fines, platforms will find it cheaper and safer to censor pro-life content that could help women than to risk ruinous penalties from the EU,” she said.

Norman Lewis, senior research fellow at Freedom and Sovereignty Initiative, told The Epoch Times by email that the issue indicates “there is either bias in the training models or bias in those designing the LLMs.”

“Either way, making it public and debating it is what free speech is all about,” he said.

Lewis, a former PwC director and former director of technology research at Orange UK, said that “demanding to know why is part of holding these AI platforms to account.”

“But arguing that the DSA, etc. (as they do in the parliamentary question) should be applied is like turkeys voting for Christmas. ChatGPT is already being constrained under the DSA. We do not need more restrictions but public accountability,” he added.

OpenAI, Google, X, and Anthropic did not respond to The Epoch Times’ request for comment.

None of the political groups representing the 68 MEPs who signed the written question responded to The Epoch Times’ request for comment.

AlgorithmWatch did not respond to The Epoch Times’ request for comment.