Nonrival

Expert analysis from think tanks and academia, on public policy, economics, and technology.
Human experts. AI summaries.

Deepfake Rules Are Easy to Write But Nearly Impossible to Enforce Before Election Day

csis

• Brazil, Israel, and Spain have all tried to regulate AI-generated political content — through outright bans, labeling requirements, or fines — but each approach has crumbled under real-world pressure, leaving judges to make case-by-case calls that often come too late.
• The core problem isn't writing the rule; it's defining the line between political satire and intentional deception, and acting fast enough to matter — deepfakes can rack up millions of views in hours, while legal challenges take weeks.
• The U.S. has no federal deepfake law, and a patchwork of state rules can't stop content that spreads across borders instantly, leaving the country poorly prepared as the 2026 midterms arrive.

Meta paid $25 billion for lying about Cambridge Analytica — but no one has ever been held accountable for the actual harm

the_conversation

• The record-breaking $25 billion judgment against Meta in New Mexico was for deceiving users about data privacy practices, not for the political manipulation that Cambridge Analytica allegedly carried out using harvested Facebook data.
• A decade of settlements — including a 2026 multistate deal that buried Cambridge Analytica claims inside a teen-safety lawsuit — has let Meta pay fines for the cover-up while the underlying harms were never litigated.
• The author, a researcher whose work helped expose the scandal, argues that investigators' failure to fully share evidence across the U.S. and U.K. left the most consequential questions unanswered, and that states like New Mexico should keep pushing.

AI is making cyberattacks faster and scarier, but most of the underlying threats are older than you think

atlantic_council

• A $6.3 billion hack of crypto exchange Binance and AI-assisted discovery of major software vulnerabilities made headlines, but the author argues these threats are more familiar than they appear — compromised credentials and crypto exchange hacks have been around for decades.
• What is genuinely new is the speed: AI tools are accelerating both the discovery of software vulnerabilities and the laundering of stolen funds, leaving defenders with less time to respond.
• The article urges regulators to distinguish real, emerging risks from AI hype before writing sweeping new rules, and notes that the same AI tools enabling attackers could help defenders patch aging software infrastructure first.

In the News

The Supreme Court Looks Ready to Let Catholic Preschools Reject LGBTQ Families and Still Get State Funding

scotusblog

• The Supreme Court heard arguments in a case where Catholic preschools in Colorado want to participate in the state's free universal preschool program without complying with its nondiscrimination rules, which require schools to admit children of LGBTQ families.
• A majority of justices appeared sympathetic to the Catholic schools, suggesting Colorado's conditions may violate the First Amendment under the same logic the court used in a 2022 ruling that struck down Maine's exclusion of religious schools from public funding.
• The case could force states to choose between funding religious schools that discriminate or scrapping publicly funded programs altogether — a potentially far-reaching consequence that Justice Sotomayor warned would come at "great cost to someone."

Analysis

Every Major Scientific Field Is Globalizing Its Top Research — Except Economics

promarket

• A study of nearly 6,000 winners of 171 major prizes across 18 fields finds that prizewinning research has spread to more institutions and countries in every field except economics, where half of all career years were spent at just eight elite American universities.
• The authors point to three culprits: economics rewards prestige signals over decisive experimental results, star economists are unusually mobile and easy to poach, and the field has an unusually narrow methodological consensus enforced through five dominant journals.
• The upshot is a warning: when governments rely on economists for policy advice, they may be drawing from an unusually narrow intellectual gene pool — a concentration that economists themselves would flag as a problem in any other market.

Also Worth a Look

• BlackRock and Vanguard Influence Companies Through Advice, Not Threats — and It Works Better on Some Issues Than Others (wharton)
• AI Doesn't Have to Make Your Decisions to Distort Them (wharton)
• Courts Swatted Down the Trump Administration's Election Interference, but the Damage to the Justice Department May Last (the_conversation)
• Arkansas Is Mandating Coverage for a Fertility Treatment Doctors Say Doesn't Work (think_global_health)
• The Supreme Court May Eliminate the Only Way to Sue Federal Officers for Constitutional Violations (scotusblog)
• Pregnant on the Border: How Pesticides, Heat, and a Lack of Hospitals Make Pregnancy in Rural Arizona Uniquely Dangerous (yale_climate)
• How Governments Use Religion as a Foreign Policy Tool — and Why the West Keeps Misreading It (atlantic_council)