New coverage from 404 Media Sam Cole highlights CDT's latest research on the manipulative "dark patterns" appearing in AI chatbots.

“The rise in humanoid companions coincides with skyrocketing cases of AI love. In the US, one in five high-schoolers has had a romantic relationship with AI or knows someone who has (Center for Democracy and Technology).”

What if AI is less like software and more like "forever chemicals?" In a new piece on her Substack, Elicitation, CDT's Miranda Bogen explores what environmental health regulation can teach us about governing AI and managing uncertain, long-term harms.

Election security depends on strong cyber info sharing. For nearly a decade, federal, state, local, & private-sector partners worked together to identify threats, coordinate responses, & strengthen defenses. But that system is now showing cracks. New from CDT's Geoff Hale:

From facial recognition to internet surveillance, governments now have unprecedented tools to collect and access personal data. Protecting security and protecting freedom are not mutually exclusive. Strong safeguards, oversight, and rights-respecting policies are essential.

What happens when AI models are fine-tuned for new tasks? CDT and MIT researchers found that even small, seemingly benign modifications can cause unpredictable “safety drift,” changing how models behave in ways developers may not expect. Check out out the research:

"This is not the time to hand over more data about ourselves to tech actors." @aliyabhatia.bsky.social Indeed.

Parents, you aren't alone in wanting to protect your kids online. Teens & parents expressed concerns about age verification in a recent survey. “As far as taking a picture with the ID, that seemed really intrusive to me.” – Parent to two teens, 16 and 13 More from @cdt.org: cdt.org/insights/tee...

In the new edition of #CODISignals, @cdt.org’s @aliyabhatia.bsky.social highlights the need for collaboration among industry, academia, civil society and others to advance multilingual AI. Read more: www.codi.global/newsroom/cod...

“Surveillance pricing ... aims to target consumers individually, to induce them to pay the highest price they are willing to pay, based on the seller’s assessment using personal data collected about them,” said @cdt.org Senior Counsel for Competition Policy George Slover.

the full paper gets into way more detail: cdt.org/insights/dar...

the one from ChatGPT is especially wild considering OpenAI started using it for "breaks" after recognizing that long sessions are actively more dangerous for its users

anthropomorphism and sycophancy are unique, new problems. but the old dark pattern tricks still apply; these popups when you try to leave, for example, are truly ridiculous

new @cdt.org study out this morning looks at how chatbots deploy manipulative "dark patterns" to get people to offer up more than they bargained for: www.404media.co/new-study-re...

Excellent new research from @cdt.org @mluria.bsky.social analyzes the dark patterns applicable to AI chatbots, including general-purpose systems (ChatGPT, Gemini, Claude) and “companion” platforms (Replika, Character.AI): cdt.org/insights/dar...

New CDT research by Ruchika Joshi, Adinawa Adjagodjou, & Michal Luria, maps “dark patterns” in AI chatbots, showing how design choices like playacting, sycophancy, and emotional framing can blur the line between helpful interaction and manipulation. cdt.org/insights/dar...

What does a 1969 conceptual art exhibit have to do with generative AI and the First Amendment? CDT’s Becca Branu explores how expression, inquiry, and creativity can persist through intermediaries from telephone lines to AI systems.

As policymakers and researchers explore insurance as a potential AI governance tool, a recent court ruling involving Meta raises difficult questions: What happens when insurers argue AI-related harms stem from intentional design choices… not accidents? Read more:

Studies from cities across the country have raised serious questions about the effectiveness of gunshot detection technology while also highlighting risks of over-policing, false alerts, and wasted public resources. CDT’s Jake Laperruque examines the reality behind the hype:

States are moving quickly to bring AI into K-12 classrooms, but guardrails matter. CDT’s Maddy Dwyer outlines 3 key policy priorities for responsible AI adoption in schools: risk management, AI governance, and transparency. Read more:

“A recent order by a DE state court siding with insurance companies in their case against Meta over coverage for the cost of lawsuits associated w/ the company’s platform design choices amplifies doubts ab the industry’s role in mitigating risks from AI, according to CDT.”

Yesterday, CDT submitted comments to the Meta Oversight Board’s consultation on governing AI chatbots for 13–17 year olds. Read the full comments:

“Becca Branum, [from CDT]'s Free Expression Project, told CyberScoop that the $53,088-per-violation penalty structure creates an incentive for platforms to remove anything that arrives through the complaint pipeline, regardless of whether it is genuine NCII.”

“School admins around the country are not prepared for ‘nudify apps & the chaos & trauma they’re creating for students. A 2024 report by the CDT found that 40% of students & 29 percent of teachers said they knew of an explicit deepfake shared in the past school year…”

“EFF, CDT, & other orgs urged a US appeals court to uphold an injunction against NSO Group Technologies from installing its spyware Pegasus on WhatsApp, arguing that it promotes the security & missions of journalists, activists, & human rights defenders in the US & world.”

“CDT’s VP of Policy Samir Jain said AI systems such as Anthropic's Claude Mythos Preview model, show the ‘offensive cyber landscape is poised to change dramatically, and small & under-resourced jurisdictions are likely to be particularly vulnerable to that change.’"

As states move to integrate AI into classrooms, transparency and accountability can’t be optional. CDT’s latest recommendations outline practical steps lawmakers can take to protect students while supporting responsible innovation in K-12 education.

Free expression online depends on more than the ability to speak — it requires the freedom to access info, participate in civic life, and make choices about what we see and share. CDT works to protect an open internet free from censorship, gatekeepers, & barriers to speech. cdt.org/area-of-focu...

Yesterday, CDT’s Samir Jain testified before the House Homeland Security Subcommittee on Cybersecurity & Infrastructure Protection on the escalating cyber threats facing state and local governments.

Can liability insurance help shape responsible AI development? A recent court decision involving Meta’s insurers raises important questions about accountability, foreseeability, and the limits of insurance as an AI governance tool. Read CDT’s analysis: cdt.org/insights/des...

AI is already transforming K-12 education, but schools need clear guardrails to protect students’ rights, privacy, and safety. CDT’s Maddy Dwyer outlines 3 key policy priorities for responsible AI adoption in schools: risk management, governance, and transparency.

HAPPENING NOW: Today at 2 PM ET, CDT’s Samir Jain is testifying before the U.S. House Homeland Security Subcommittee on Cybersecurity & Infrastructure Protection on the escalating cyber threats facing state and local governments. homeland.house.gov/h...

CDT’s Kate Ruane in the Associated Press on the free speech concerns of livestream platforms: “It’s not clear what was done to improve that situation before [the functionality] was reinstated.”

74% of Americans are concerned about the privacy and security of personal data held by government agencies, & nearly 8 in 10 believe Congress should hold agencies accountable when privacy laws are ignored. Join CDT on June 1:

Tomorrow at 2 PM ET, CDT’s Samir Jain will testify before the U.S. House Homeland Security Subcommittee on Cybersecurity & Infrastructure Protection about the escalating cyber threats facing state and local governments. 📺 Watch live: homeland.house.gov/h...

“According to a recent survey by the nonpartisan CDT, 57% of special education teachers polled nationwide said they used AI to help develop individualized plans for their students in the 2024-25 school year. That's up from 39% the previous school year.” www.npr.org/2026/05/...

A recent court decision denying Meta insurance coverage in social media harm litigation could have major implications for AI governance. New analysis from Miranda Bogen & Tina Park explores what the case reveals ab the limits of insurance as a tool for AI accountability.

“A report by CDT found that a similar percentage of teachers used AI tools in the 2024-2025 school year. Less than 4 years after the launch of ChatGPT, AI is changing how students learn in profound and sometimes troubling ways that research is only just starting to document.”

Canadians overwhelmingly support encryption and reject expanded government access to private messages. New research by CDT finds deep public concern with key surveillance powers proposed in Bill C-22, including risks to privacy, free expression, and online security.

“Becca Branum, director of the Free Expression Project at CDT, told CyberScoop that some elements of the FTC’s approach – like requiring clear and simple reporting options for victims – aligns with best practices established by civil society groups.”

“An April 30 report from NIST partner Center for Democracy and Technology (CDT) noted the need for broader engagement across the standards community to help translate contextual evaluations into real-world applications.”

CDT’s Jake Laperruque hosted a new Tech Policy Press podcast unpacking an important case before SCOTUS this term: Chatrie v. United States. At issue are geofence warrants, which allow police to demand data on every cellphone in a specific place & time.

ICYMI: CDT’s Aliya Bhatia wrote in The Guardian about the growing wave of age verification laws that could force people to upload government IDs just to access everyday online platforms. The piece explores what’s at stake as govts push new restrictions.

CDT's Hannah Quay-de la Vallee and Quinn Anex-Ries break down how the Pentagon's dispute with Anthropic isn't just a defense issue. It's disrupting AI adoption across every level of government, from federal agencies to state, local, and tribal governments.

“...Elizabeth Laird, director of equity in civic technology at CDT, described the education sector as a ‘target-rich environment,’ as schools and vendors collect highly sensitive information about tens of millions of Americans, including minors.”

How should the government protect your personal data, and who's holding agencies accountable when they don't? Join CDT June 1 (12–1pm ET) for a virtual panel on government data privacy with experts from civil society and government agencies. Register now: cdt-org.zoom.us/webi...

CDT’s Travis Hall in Governing: “[Colorado’s SB 189 bill is] no longer really a discrimination bill. It is now almost entirely a disclosure-focused bill.”

“A 2025 report by CDT also shows that an average of 71% of K-12 teachers reported that when students use AI to complete their schoolwork, it is hard for the teachers to understand whether student work is their own.”

🎉 I've joined @cdt.org as their Research Director. 🎉 I could not be more excited to be part of an organization that has championed people's online rights, fought for good technologies, and defended against problematic ones for decades. If you have research ideas, reach out!

CDT is one of Humanity AI’s inaugural group of grantees! Humanity AI is supporting a diverse group of organizations committed to putting people at the center of AI development.