Reproductive Justice in the Age of AI

/

In March of 2024, at the University of Cincinnati’s Black Feminist Symposium, I facilitated a workshop called AI and Reproductive Rights: Co-Creating an Inclusive Future. The room held students, professors, technologists, and people who had never touched a line of code. When we reached the example of Amazon’s abandoned hiring algorithm— trained on ten years of resumes until it penalized any resume that mentioned women — what surprised people was the date. This had already happened, years ago, and we were only now catching up to it.

The story landed differently in 2024 than it would have a few years earlier. Words like bias, feedback loop, and systemic harm had become ordinary by then, part of how communities talked about AI in general. Nobody in that room needed them explained. But almost none of that language was in circulation when Amazon built and scrapped the algorithm back in 2015. That was the part that stuck with people. The harm had existed for years before there was public vocabulary that could hold it. The women whose resumes were filtered out at the time had no mainstream vocabulary for what was happening to them. That delay between harm and recognition is where technology settles into everyday life before we learn how to argue with it. And by the time the argument is ready, the technology is no longer optional.

That same delay is shaping reproductive health now. Automation has been part of this landscape for years. AI is the newer layer, and it enters a field already structured by legal uncertainty, charged political messaging, and uneven access to reliable information.

Misinformation in an automated ecosystem

In Ohio, lawmakers recently advanced a proposal to require public schools to show Meet Baby Olivia, a computer-generated video about fetal development produced by the anti-abortion group Live Action. This kind of policy does more than express ideology. It shapes what feels factual,  what feels neutral, and what feels medically settled before people have the language to contest it. 

Layered on top of that, the landscape of abortion law, by design, is almost impossible to follow in real time. Rules vary state to state, change every time a court rules on them, and shift again depending on how prosecutors decide to enforce them. For anyone trying to figure out what’s legal where they live, none of this is easy to track. They piece it together from wherever they can find it: a friend, a search result, a news article, a hotline. That’s the environment AI walks into. When someone types a question into a chatbot and gets an answer back, they read it as more than information. Whether or not they came looking for advice, that’s what they walk away with.

Legal advocacy organizations documented what happened when they asked AI systems how to get an abortion in Texas. ChatGPT referred users to anti-abortion crisis pregnancy centers. Google’s Gemini failed to mention that abortion pills can still be obtained through shield-state providers. Campaign for Accountability tested multiple AI answer engines and found repeated referrals to a hotline promoting “abortion pill reversal,” a practice that major medical bodies describe as unproven and potentially dangerous.

The problem runs deeper than one flawed model. It shows how automation absorbs the loudest material in an information ecosystem and redistributes it with institutional authority. Abortion disinformation already circulates through paid advertising, crisis pregnancy centers, restricted sex education, and politically mandated media. AI becomes another part of that structure, and it gives that misinformation credibility.

The person typing the question in most cases is too vulnerable to vulnerable to stop and verify. They get one confident answer and act on it. 

Data markets as enforcement infrastructure

Even accurate information has consequences. Seeking it produces records.

That was true long before the fall of Roe. What changed afterward is the meaning of those records. They became politically actionable.

In 2024, Senator Ron Wyden disclosed that a location data broker, Near Intelligence, tracked  visits to nearly 600 Planned Parenthood locations and that this data helped support targeted anti abortion advertising. Two years earlier, the Federal Trade Commission sued Kochava for selling  precise geolocation data that could reveal visits to sensitive locations, including reproductive  health clinics. 

This is what surveillance looks like in practice. These systems were built to collect and sell location data at scale, and reaching abortion patients is not a new use of them. It is the same use. Tracking people is the product, and that is why privacy here cannot be a matter of personal caution. The same systems that optimize advertising make intimate decisions legible to buyers.

The response to this cannot be personal. The capacity was assembled at the level of collection and exchange, and that is the only level at which it can be contested.

Nonprofit reality

In my day-to-day role at Planned Parenthood, I arrived with a technical background and an assumption that nonprofit health work would sit at a distance from the culture of rapid automation. It does move differently. It is slower, and more careful. It is also not outside of it.

I have seen members of my team and colleagues across affiliates reach for AI tools because nonprofit work is always negotiating capacity. I use them myself to troubleshoot code while cleaning data, with strict boundaries and no identifiers, because the work still has to get done. None of this is enthusiasm for the tools. It is what happens when institutions are asked to absorb the demands of scale while actively being stripped of their resources, and the people inside them close that gap with whatever is at hand.

I see the other side of this too. An older community member trying to reach us happened to get my work number and spent the first full minute of the call making sure I was a real person before they would say why they had called. It is hard to blame them. The assumption that a voice on the line belongs to a real person has stopped being something you can take for granted. That is the strange ground we are doing care work on.

This is why refusal cannot be our policy model. Stepping back does not pull automation out of healthcare, legal navigation, or public communication. It only hands those defaults to the people who set them for profit.

Which brings me back to where I started. At that workshop I ran at the Black Feminist Symposium, we spent the last hour futurecasting. Everyone picked a reproductive justice problem they cared about and sketched a technical response to it, starting from the consent, privacy, and safety of the person who would be affected, before anything was tuned for speed or scale. People who had opened the day saying they did not build technology spent that hour designing it. They stopped treating these systems as fixed conditions and started treating them as choices they could have a say in.

That refusal is the tradition they were stepping into, whether they knew it or not. Reproductive justice was built by Black feminists who never accepted the conditions they were handed as final, who treated the given world as something that could be organized and rewritten. The instinct in that room was not new. It was the latest turn of very old work.

AI is shaping reproductive life, and the lives it shapes are the ones who should get to imagine and design it.

Sources

Author Bio: Aashka Raval is a Data Analyst on the Development team at Planned Parenthood Southwest Ohio and a University of Cincinnati alum. They are a former Paragon Policy Fellow, where they conducted cybersecurity policy research for the City of Bismarck. Their work sits at the intersection of technology, reproductive justice, and public interest data systems.