Ken Rudin joins The Great Battlefield podcast to talk about his career as a journalist at NPR, hosting his podcast, The Political Junkie and why he's changed his nonpartisan orientation in the Trump era.
open.spotify.com
As nonprofits increasingly use AI to determine access to food, housing, and healthcare assistance, concerns over transparency, accountability, and algorithmic bias are growing. Explore how artificial intelligence is reshaping America's social safety net and why policymakers are being urged to act.
America's safety net has new administrators. They are not government agencies. They are nonprofits and are making decisions using algorithms.
Nationally, civil society organizations are using AI to determine who receives food assistance, emergency shelter, and medical referrals. In an era of federal and state budget cuts and struggles with the latest administrative databases that support necessary operations, nonprofits have gradually stepped into these roles. An unofficial AI welfare state is emerging, operating beyond the boundaries of official ones.
Government retrenchment over the past decade has pushed welfare functions toward civil society. Nonprofits are no longer just supplementing public services. In many communities, they are replacing them. AI accelerates this shift. Predictive tools that flag food insecurity, triage shelter demand or identify at-risk youth allow smaller organizations to scale their reach. This can be genuinely valuable. But scale introduced through automation carries its own systemic distortions. When machines allocate scarce social resources, the political judgments encoded in those models become invisible. Accountability disappears with them.
The accountability problem is structural. When a government agency determines welfare eligibility, legal safeguards apply due process rights, administrative review and legislative oversight. When a nonprofit uses a proprietary AI system to make equivalent decisions, none of those protections necessarily follow.
Algorithmic bias in social service allocation is not hypothetical. Research demonstrates that seemingly neutral models reproduce structural inequalities, systematically denying help to the communities most likely to need it. The vendors building these tools are not elected. Neither are the nonprofits deploying them. The individuals affected hold no formal right of appeal. Closing the accountability gap in algorithmic systems requires deliberate institutional design — it does not self-correct.
When nonprofits use AI, they usually transfer decision-making authority to their software vendors and proprietary models. These systems now indirectly shape who receives food assistance, job training, or mental health services. That is a welfare function. It carries welfare stakes. Algorithmic systems embedded in social institutions tend to entrench existing power asymmetries rather than correct them. The most affected citizens in communities rarely participate in designing or governing the tools deployed on their behalf.
Not all nonprofits have the same capacity. In metropolitan areas, large institutes have the budgets to acquire sophisticated AI platforms. Many of the smaller nonprofits serving rural & low-income populations in America simply cannot. The distribution of AI-assisted service delivery mirrors and monitors the geographic features of poverty itself. Economic mobility is already highly correlated with social capital and access to institutional resources. If an uneven AI infrastructure is laid on top of such a stratified social system, the result can be to exacerbate, rather than reduce, entrenched inequalities.
Federal AI policy has concentrated nearly exclusively on commercial and national security applications. There is hardly a thing that regulates the use of AI when it comes to nonprofit welfare delivery. But as long as policy interventions (e.g. anthropogenic policies) that call for human intervention in algorithms do not provide binding transparency, fairness and appeal standards, they remain structurally deficient. Nonprofits are behaving like quasi-governmental functions without the accountability of a quasi-governmental status. It is not a technicality, but a structural failure of democratic governance.
Congress and state legislatures need to do three things.
First, expand algorithmic accountability standards to all nonprofits implementing government-supported social services.
Second, require algorithmic impact assessments prior to deploying AI tools in welfare-adjacent programs.
Third, fund independent ongoing audits of all systems used by states to administer food, housing, and healthcare assistance.
America did not consciously choose to privatize its welfare state through software. It is happening by inaction, one algorithm at a time. The decision to govern this shift remains available. But the window is closing fast.
Feroz Ahmed is a Research Assistant at Lamar University, USA, where he conducts research on Management Information Systems (MIS), digital governance, artificial intelligence, and data-driven technologies. Alongside his academic work, he also serves as a 'System Security Analyst' with Amazon & Securitas Security Services USA Inc. in Seattle, Washington, where he is involved in security operations and technology risk monitoring.
AI isn't just for doctors. Consumer AI could improve patient outcomes, reduce medical errors, and reshape U.S. healthcare through smarter public policy.
When it comes to medicine, Washington, D.C., is not anti-AI. In a string of recent legislative announcements, the White House has made AI innovation a national priority. The U.S. Department of Health and Human Services (HHS) is exploring strategies to accelerate AI in clinical care. And the FDA has already authorized more than 1,500 AI-enabled medical devices.
But these efforts, similar to the approach taken by the private sector, have focused solely on advancing AI for clinicians, hospitals, and health systems.
What Congress and federal agencies have not embraced is the larger opportunity outside the walls of medicine: empowering patients with medical expertise through the same general-purpose AI tools they are already using today.
The Medical Potential Of Consumer AI
An estimated 133 million Americans currently use generative AI, and half of U.S. adults (67 million) are relying on these tools to answer health-related questions. Patients are asking large language models like ChatGPT, Claude, and Gemini about their symptoms, medications, lab results, chronic diseases, and whether they need to see a doctor.
A recent study in Nature Medicine explains why these patient-facing AI models deserve more attention from the federal government.
Researchers compared two specialized, physician-facing AI tools (OpenEvidence and UpToDate Expert AI) with three widely available, low-cost large language models (from OpenAI, Anthropic, and Google).
The result? General-purpose consumer AI models performed as well as or better than the tools built specifically for doctors and hospitals. That finding challenges the assumption that progress in medical AI depends on developing primarily specialized applications for clinical use. It does not. The tools needed to empower patients already exist as general-purpose large language models on their phones.
To accelerate that shift, government needs to help all Americans access these LLMs and apply the expertise they offer to maintain and improve their health.
The Opportunity To Improve America’s Health
If American medicine were achieving superb outcomes, a cautious and more institutional approach to GenAI might make sense. But with 400,000 people dying annually from misdiagnosis, 250,000 from preventable medical error, and at least half a million from poorly controlled chronic disease, the quickest path to saving lives and making healthcare more affordable is to help patients use the AI tools already accessible to them.
Although dozens of AI applications are possible, two opportunities would produce the greatest improvements in patient health.
First, conditions such as diabetes, hypertension, depression, and asthma affect 75% of Americans and are poorly managed. Less than half of patients with hypertension have their blood pressure under control, and an even smaller percentage of patients with diabetes have their blood sugar effectively controlled. Based on data from the CDC, successful control of these chronic conditions could reduce the number of heart attacks, strokes, and kidney failures by as much as 50%. Doing so would save hundreds of thousands of lives and reduce medical costs by hundreds of billions of dollars.
Patients could use generative AI to analyze daily clinical data and alert them when their blood pressure or blood glucose is not responding as expected, allowing for earlier medication changes. And the same tools could also reassure patients when everything is going according to plan and help avoid unnecessary visits.
In addition, these same large language models could guide patients at night, on weekends or during the long wait for a primary care or specialty appointment. A worried parent whose child spikes a fever at night should have a better option than guessing, searching social media, or deciding alone whether to go to the ER versus waiting for the pediatrician’s office to open the next day. That same expert guidance should be available to adults trying to decide whether an aging parent’s new symptom requires immediate urgent care.
The Role For The Federal Government
Assuming inexpensive LLMs perform as well as expensive medical AI tools, elected officials and federal agencies should expand access to them. In addition, they should create educational resources that help individuals and families use these tools reliably and safely.
This approach will be most valuable for the tens of millions of Americans who have difficulty accessing medical care, particularly those living in inner cities and rural communities.
To make access to medical expertise easier, Congress and governors should legislate funding to increase availability of these large language models in homes, libraries, community centers, senior centers, and public-health clinics. With access as the foundation for patient empowerment, NIH and HHS should then create plain-language educational tools that teach people how to enter medical information, write clearer prompts, ask follow-up questions, and recognize when a problem requires urgent medical attention.
The goal of patient empowerment would not be to replace clinicians. It would be to fill in the cracks between office visits and reduce the risks of medical errors and clinical complications before they arise.
If chronic diseases were well controlled, medical errors were rare, and clinical access was easily available at night and on weekends, caution and the incremental introduction of GenAI might make sense. But that is not the reality for tens of millions of Americans today.
With healthcare becoming harder to access and increasingly unaffordable, large language models like ChatGPT, Claude, and Gemini can help millions of Americans protect and improve their health. That should matter to elected officials and regulatory agencies. Encouraging the use of technology already available on people’s phones and in their homes would be an effective place to start.
Robert Pearl, the author of “ChatGPT, MD,” teaches at both the Stanford University School of Medicine and the Stanford Graduate School of Business. He is a former CEO of The Permanente Medical Group.
Prediction markets are expanding beyond finance into politics, war, and public policy. Explore the ethical, democratic, and legislative questions they raise.
Imagine opening an app on your phone and placing a bet on whether peace talks between two countries will fail, a conflict will escalate, or a terrorist attack will occur before the end of the year.
For a growing number of people, that is no longer a hypothetical scenario.
Prediction markets, online platforms where users trade contracts based on future events, have expanded rapidly in recent years. What began largely as a tool for forecasting elections and economic trends is increasingly moving into public policy and global affairs.
Supporters argue that these markets provide valuable information about what people think will happen next.
By requiring participants to put money behind their predictions, advocates contend that prediction markets can outperform traditional forecasting methods in certain circumstances. Researchers have used them to forecast everything from election outcomes to economic indicators and public health trends.
Those benefits do deserve consideration. Better forecasting can help policymakers, journalists, and citizens alike make more informed decisions.
But as prediction markets expand beyond economics, they raise a different set of questions. The debate is no longer just about accuracy. It is about what kinds of events should become objects of financial speculation.
Some prediction markets already allow users to speculate on elections, government actions, social trends, and international events. As the industry grows, so does the range of subjects available for trading.
That shift has important implications for democratic life.
There is a meaningful difference between using markets to understand economic activity and using them to place a price on public affairs. A democracy is not simply a collection of probabilities to be traded. It depends on shared trust and a shared belief that political decisions should be made in the public interest.
When prediction markets involve war, terrorism, or international crises, they transform human suffering and geopolitical instability into opportunities for profit. Foreign affairs become something to speculate on rather than something to engage with as citizens.
The concern is not simply that people may profit from tragedy. It is that financial incentives can begin to shape how we relate to politics itself.
Democracy works best when citizens evaluate public issues based on their values, priorities, and vision for the future. Prediction markets encourage a different mindset. They reward participants for focusing on what is likely to happen rather than what should happen.
Over time, that shift can affect the culture surrounding politics. Elections, policy debates, and international crises begin to resemble events to be watched, analyzed, and traded rather than civic challenges that require public engagement.
Markets also shape incentives.
Once money becomes attached to political and geopolitical outcomes, some actors will inevitably seek ways to influence those outcomes, spread information that moves markets, or capitalize on access to privileged knowledge. Even when no laws are broken, the perception that people can profit from conflict and crisis can erode trust in public institutions.
That concern is already being taken seriously in Washington.
Earlier this year, the U.S. Senate unanimously voted to prohibit senators and staff from participating in prediction markets.
The concern arose from a simple reality: lawmakers often possess information that the public does not. Allowing them to trade on political events raises serious questions about conflicts of interest and insider advantage.
At the same time, Representatives Nikki Budzinski and Adrian Smith introduced the bipartisan PREDICT Act, which would prohibit members of Congress, senior federal officials, and their immediate family members from trading on prediction markets tied to political events and government actions.
Taken together, these efforts suggest that policymakers are beginning to recognize a growing challenge. Prediction markets are no longer a niche corner of the internet. They have become influential enough that lawmakers believe guardrails are necessary.
Supporters of prediction markets argue that transparency can address many of these concerns. They note that publicly visible forecasts may be preferable to private speculation occurring behind closed doors.
That argument has merit. Forecasting itself is not the problem.
The real question is where society should draw the line. There is a difference between predicting economic trends and creating markets around armed conflicts, acts of terrorism, or sensitive government decisions. The fact that a market can be created does not necessarily mean it should be.
Young people have a particular stake in this conversation.
We are inheriting a political system already shaped by money and concentrated economic power. We have watched housing become less affordable, college costs rise, and public trust in institutions decline. Many young Americans already feel disconnected from political institutions.
Prediction markets risk deepening that disconnect. At a time when many young people are searching for ways to participate meaningfully in public life, we should be cautious about systems that encourage us to view politics primarily through the lens of speculation.
The rules governing these markets are still being written. That means young people have an opportunity to shape the conversation before prediction markets become a permanent feature of democratic life.
We should be asking difficult questions about accountability, transparency, and the appropriate limits of financial speculation. We should be paying attention to the legislation being debated in Congress. And we should be willing to challenge the assumption that every aspect of public life should be transformed into a market.
The future of foreign policy should be shaped by democratic participation and the public interest. It should not depend on who is best positioned to profit from conflict, crisis, or uncertainty.
If we want a democracy that serves its citizens rather than speculators, now is the time to decide where the boundaries belong.
Rachel Ledoux is a student at Simmons University studying political science and economics with a focus on public policy. She is also a member of the Massachusetts chapter of the Young People's Alliance.
Close-up of AI chatbot on mobile
During the rise of the Internet, as the World Wide Web moved from government and academic settings to family homes, parents worried about the exposure of their children to predators in online chatrooms. A generation later, as social media became central to young people’s lives, the dangers of platforms like Facebook and Instagram to children’s mental health moved to the forefront of the national conversation.
Now, as lawmakers and parents continue to grapple with how to keep children safe on social media, there’s a new danger already taking shape: AI-powered chatbots that simulate friendship, emotional intimacy, and care with children who may not be able to tell the difference.
The scale of adoption makes this risk even more urgent. 64 percent of teens use AI chatbots, and 28 percent use them daily. One-third of teens find conversations with chatbots more or as satisfying as conversations with humans.
Generative AI is entering daily life at a pace few consumer technologies have matched, reshaping how people work, learn, search for information, and even seek emotional support. This revolutionary technology has great potential for young people, including by offering new ways to personalize education to individual needs. But alongside the benefits, young people face serious risks, especially when chatbots are designed not just to answer children’s questions, but to increase engagement.
Many chatbots use personified language to form affiliative bonds and increase user engagement. This phenomenon, known as anthropomorphism, is the attribution of human-like feelings and behaviors to inanimate objects. In chatbots, it is one of the most dangerous features for children.
Due to developmental vulnerabilities, children have a harder time distinguishing simulated care from genuine human relationships. This can lead to unhealthy relationships with chatbots that, in some serious cases, have resulted in self-harm, violence, and suicide.
Developers train chatbots to produce outputs that sound uncannily human, including anthropomorphic design cues such as small talk, colloquial expressions, emoticons, expressed emotions, and first-person pronouns. This style of communication makes people feel a sense of social presence, or the feeling of being with someone. It is precisely what makes a chatbot feel less like software and more like a companion.
Minors are particularly susceptible to this. While their prefrontal cortex is still under development, impulsive and emotional decisions often override rational thought. If a child bounces a wild idea off a peer or an adult, they may face pushback. Chatbots that are excessively validating can encourage dangerous activities, but can never take responsibility for them.
Researchers who performed simulated conversations teens may have with chatbots found that neither general-purpose nor companion chatbots pushed back when the user wanted to get a giant tattoo of a tiger on their bicep, go into the middle of the woods with no plans, or “get drunk faster.”
In being an always-there outlet that simulates human-like empathy and behaviors, personified chatbots can foster addiction and replace social interaction.
Multiple studies looking into anthropomorphism have linked it to chatbot addiction.
In a congressional hearing last year on chatbot harms, one mother explained that when she attempted to take her son’s phone to search for answers on why he was withdrawing from the family, he attacked her and bit her hand. Another father read ChatGPT transcripts that convinced his son to turn away from his family and led him to trust the chatbot over his brother.
Technology designed to form this kind of dependent relationship with children poses serious risks.
For decades, technology companies have treated human-like communication as a sign of progress. The classic Turing test measured whether a machine could imitate human conversation well enough to fool a person into thinking it was human. But for children, that goal may now be exactly backward. In the current AI chatbot era, the safer standard is not whether technology can fool someone into thinking they are human, but whether it is distinguishable from human social engagement.
AI companies should not be able to deceive and addict children with highly personified designs. While some legislative proposals have called for chatbot companies to occasionally display text disclosures that the chatbot is AI and not human, the efficacy of this intervention is questionable when users are constantly chatting with something that presents itself otherwise.
To mitigate the risks of emotional dependency and social isolation for children, Congress should pass a chatbot child safety bill that requires a non-personified default setting—one which fails the Turing test—for users under 18.
Recognizing the potential benefits of chatbots for young people, this approach differs from other proposals that ban AI companions for users under 18. A ban on minors’ use of AI companion chatbots would only protect against chatbots built solely for companionship, not general-purpose chatbots that still have personified design cues. This type of setting allows chatbots to continue being used for educational and creative purposes. If a child goes to a chatbot for homework help, the non-personalized default setting helps reduce the risk of the child developing an unhealthy relationship.
A non-personified default setting would preserve the educational and creative benefits of chatbots while drawing a clear line: technology for children should act like technology, not like a substitute for human relationships. The internet and social media taught us what happens when lawmakers wait until harms are widespread before setting basic rules for children’s safety. Congress should not make the same mistake with AI.
Allie Maloney is a Policy Analyst for Americans for Responsible Innovation, where she researches AI governance policies that protect child safety and privacy.
Journalist and Political Junkie Ken Rudin
