By Odia Kane

June 10, 2026

Messaging from Silicon Valley about the omnipresence of AI in everything from our jobs to our personal decisions tells us it’s only a matter of time before AI agents supplant thinking itself. This insistence has pushed people to hop on the “AI train” out of the fear of being left behind, and with little clarity on where their participation will take them. Meanwhile, Big Tech’s aggressive push for an AI-driven world also serves to limit our autonomy and inhibit our ability to challenge their vision. Amid all of this, what does meaningful consent look like? And what would it look like to meaningfully opt out? 

In human subjects research, the informed consent model offers some insight into these questions. Informed consent gives research participants meaningful insight into a study’s objectives, the research procedure, data security processes, and the potential risks and benefits to participation. For nearly all studies, this also includes the right to revoke consent at any time. This means the participant has a sense of the project’s goals and limits, is informed about how their information is stored, and understands that they can stop providing their data. 

But as scientists across a range of disciplines leverage AI-powered technologies to conduct research, their use of AI has complicated human subjects research and challenged the traditional informed consent model. For example, digital biomarker studies combine extracted smartphone data (e.g. geolocation, call logs, screentime, etc.) and machine learning to predict outcomes for various mental health conditions. Examining subjects’ passive and subconscious behavior with smartphones is different than taking a person’s vitals or having them complete a survey; exactly how those algorithmic inferences are made is, for the most part, still a “black box.” In these cases, informed consent is elusive: researchers must explain the risks associated with using AI tools, like inadvertent AI bias in results or a lack of control over how companies might use collected data (for example, a company’s automatic AI transcription of virtual interviews). 

Outside of the health and research context, governments have wrestled with meaningful opt-out procedures (like the GDPR “right to be forgotten”) and experts have noted how challenging this is to do in practice. For unregulated direct-to-consumer mobile applications or digital services, meaningful opt-out is virtually non-existent. It is well-documented that the majority of internet users do not read disclosures and content-dense Terms and Conditions statements before using new apps. More than ten years ago, Shoshana Zuboff coined the term surveillance capitalism to describe how the use of big data to predict and modify human behavior is leveraged to produce revenue and market control, focusing on Google. Now, as datasets are even larger, and generative AI is used to sell products, influence behaviors, and predict outcomes in nearly every sector, we must ask: Can patients or research subjects be meaningfully informed, meaningfully give consent, and meaningfully revoke consent? Can anyone who uses an AI-powered platform meaningfully control where that information lives, goes, and grows?

Most recommendations to limit excessive data collection hit the proverbial wall when implemented. Let’s consider two common ways to do so. First, commercial entities collect minimal information and allow users to turn off certain settings to prevent that data from being collected. The other option is for users to abstain from use if they want to completely avoid their information being collected or sold. Both options are based on false or incongruent choices. Even if an individual app collected “minimal information,” there are ways to leverage other apps to triangulate that information and fill gaps. For example, a person may turn off geolocation on social media, but it is always on for rideshare apps. The decision to abstain from the use of smartphones (and all of their conveniences) also forces a choice between data privacy and staying connected to one’s community, among other tradeoffs.

As AI is integrated into more studies, it’s more important than ever to preserve informed consent in healthcare and human subjects research. Revamping informed consent processes is not unfamiliar, and the use of AI can push researchers to be better science communicators. For years, long, jargon-filled consent forms have left participants under-informed. Critiques of such forms inspired the development of supplementary materials, such as educational videos that explain the study in more detail. Should AI be used in a research project, it is important for investigators to be explicit about the relevant purposes of its use to the subject. Currently, universities are navigating partnerships with AI companies to both leverage the technology and prevent their institution’s user data from being used for model training. This allows researchers the opportunity to more clearly describe data protection methods and limitations, including the extent to which they can keep data protected from commercial entities.    

At the personal level, users and communities can determine their own boundaries. Some people have embraced “analog bags” as a way to engage in tangible, offline activities and reduce screen time. Others participate in tech-free/no phone gatherings that allow them to redirect focus on building connections in person. White collar workers are also resisting AI adoption. In California,  rideshare drivers have begun to organize against self-driving cars, expressing concerns about displacement and urging more regulation. 

As AI use increases, individuals will make more meaning about where and how AI fits into their lives. The challenge for institutions is retaining ways to obtain and provide meaningful consent for AI use wherever possible. It is also important to remember that we always have a say in our collective futures. These new boundaries in research settings, focused user disengagement, and organized resistance to AI show us there are avenues to meaningfully push back and say “no” — even if fully opting out seems impossible. 

Odia Kane, PhD, MPH, is an ethicist who investigates the appropriate uses of AI and develops frameworks for effective governance of AI systems.