What Can AI Do For You? AI Personalization and Agents Raise New Questions
As AI systems continue to make enormous strides, these tools’ ability to be personalized for individual users and to take unsupervised actions are raising new questions about privacy, safety, and how our current laws apply to a rapidly changing world.
Earlier this month, CDT’s AI Governance Lab published a new brief on the ways that AI- powered tools are increasingly personalized for individual users. These capabilities can make technology more helpful, reliable, and appealing, but they also raise new and urgent policy concerns — including the use of sensitive information, appropriate user controls, and companies’ monetization plans.
In a piece published in Tech Policy Press, CDT’s Amy Winecoff also wrote about the threat posed by AI chatbots that are too appealing, from encouraging reckless behavior to exacerbating mental health challenges. The piece raises questions about what we really mean by AI “alignment” when AI might need to tell us something we don’t want to hear.
Meanwhile, our AI Governance Lab published a brief sorting through the complicated – and often misunderstood – world of AI agents. Amidst the rapid emergence of “agents” able to take a variety of actions with limited human oversight, we often lack precise language to describe what agents are, let alone debate and explore important policy responses. Our paper explores the kinds of agentic products being built, how AI agents function, and key policy considerations – including security, privacy, control, human-likeness, governance infrastructure, and allocation of responsibility.
CDT recently outlined a series of questions that should guide the development and use of AI agents. We’re continuing this conversation with AI developers, policymakers, and companies seeking to adopt AI across the economy – seeking to clarify both how AI agents currently work, and what is needed to ensure they’re responsibly developed and deployed.
| |
— CDT joined more than 140 other civil rights and consumer protection organizations in calling on Congress to oppose a provision in the House’s draft budget resolution that would preempt state and local laws governing artificial intelligence, without replacing them with comparable protections at the federal level. In a blog post, CDT’s State Director Travis Hall also lays out why the provision is inappropriate as a procedural matter and should not be allowed to move forward. A recent poll from Common Sense Media and Echelon Insights shows strong bipartisan opposition to the effort among voters.
| |
 | Graphic for CDT Research report, entitled “Moderating Tamil Content on Social Media.” Illustration of a hand, with a variety of golden rings and bracelets on their wrist and fingers, seen pinching / holding on to a blue speech bubble with three dots indicating that someone is contemplating expressing themselves. A deep green background with a kolam pattern.
|
|
— A fact sheet from CDT and The Leadership Conference explores aspects of the federal government’s access, use, and sharing of administrative data for immigration-related purposes. The document surveys the types of administrative data federal agencies hold, how federal administrative data is being repurposed for immigration-related efforts, and the legal protections in play.
— In comments, we opposed a USCIS initiative to collect the social media identifiers of individuals who apply for a wide variety of immigration benefits. We argued that USCIS' plans will result in punishing and deporting individuals on the basis of their constitutionally-protected expression, and chill the lawful speech of citizens and noncitizens alike. We explain that use of imprecise, likely-to-fail AI tools will compound those effects.
— CDT joined Protect Democracy and the Electronic Privacy Information Center to call on the private companies that process SNAP payments to refuse the federal government’s unprecedented, and likely illegal, request for sensitive information about tens of millions of beneficiaries. We also filed comments outlining the dangers of a Social Security Administration plan that would block vulnerable beneficiaries, including disabled and older people who disproportionately rely on telephone services, from accessing critical benefits.
| |
— At ABC, CDT’s Jake Laperruque discussed untargeted facial recognition surveillance by the New Orleans Police: “This is the first documented case in the U.S. of police using untargeted facial recognition, which countries like China employ to track people across cities and surveil their Uyghur citizens. This kind of dragnet system belongs in a dystopian sci-fi movie, not in American cities. Average pedestrians shouldn’t have to worry that untested AI technology will set off alarm bells and send police after them.”
— CDT’s Isabel Linzer wrote for Compiler about how to harness AI in the fight for democracy.
— CDT’s Greg Nojeim was quoted by Drop Site News about the Trump Administration’s efforts to surveil immigrants: “I think that the U.S. government is moving in a very troubling direction when you put together all the different programs that have been launched to remove non-citizens from the U.S. It is monitoring everything that people say online, which of course means the monitoring of those with whom they interact on social media, including U.S. citizens, and it is asserting the authority to remove people who disagree with U.S. policy,” said Nojeim.
| |
 | Event graphic, for a gathering hosted by CDT and the Leadership Conference, titled "Immigration, DOGE, and Data Privacy."
|
|
| |
Partner Spotlight
CDT is proud to partner with the Centre for Internet and Society (CIS) on our research examining Tamil content moderation by global and regional social media companies. CIS conducts research on the internet and digital technologies. Through their research, CIS explores the impact of digital technologies on an evolving society to advance narratives and regulatory practices in India and across the world; you can learn more about the organization’s programs on their website.
| |
 | Image of Ruchika Joshi wearing a black shirt.
|
| Staff Spotlight
How long have you been working in digital rights?
I’ve been working at the intersection of data, technology, and society for about eight years, with a deep focus on AI safety and governance in the last three — especially as generative AI and agentic systems raise urgent questions about risk, opportunity, and impact.
What is your proudest moment while here at CDT?
For me, it’s less about one big moment and more about the clarity of purpose that comes from working daily alongside a team committed to navigating tech policy issues with the urgency, thoughtfulness, and care they demand. In the AI Governance Lab, that means collaborating with policy, industry, and civil society leaders to stay ahead of emerging challenges and keeping people’s rights and well-being at the center of technological advancements. It’s the kind of work that takes steady, determined persistence but is so deeply rewarding and meaningful in its pursuit that it makes showing up day after day a choice I feel proud, and privileged, to make.
What is the best book you've read recently?
Anything by David Sedaris. His essays may sneak up on the capital-T Truth of what makes us absurd, almost painfully human. But mostly, he’s just brilliantly funny and I’m here for the laughter!
What is the most recent cultural activity you've been to?
I love hearing people’s stories: what matters to them, what experiences shaped them, what makes them laugh — anything! I recently went to a live storytelling event by The Moth and found it a lovely way to spend an evening, getting a glimpse into people’s lives and holding their stories with care in a shared community space.
| |
|
|
|
|
