Nonrival

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

Democrats Have Flipped: They're Now More Worried Than Republicans About AI Taking Jobs

pew

• For the first time, Democrats are more concerned than Republicans about AI's role in daily life — 56% of Democrats say they're more worried than excited, compared to about half of Republicans.
• The shift is driven by liberal Democrats, whose concern jumped from 45% to 63% since 2023, while conservative Republicans became notably less worried over the same period.
• Democrats now also outpace Republicans in expecting AI to destroy jobs over the next 20 years (75% vs. 68%), a complete reversal from two years ago.

AI Security Threats Are Real, But the Doomsday Panic Is Overblown

csis

• The most alarming finding in Anthropic's recent misuse report isn't that sophisticated state actors like China and Iran are using AI — they were already doing what AI helps them do, just slower. The real concern is that AI is giving dangerous capability upgrades to previously unsophisticated actors like Houthi rebels and hacktivists.
• The public alarm around AI threats is partly a product of tech companies stumbling into threat intelligence without the geopolitical and strategic context that national security professionals bring, leading to decontextualized disclosures that generate fear without clarity.
• The author argues the national security community needs to step up its public communication — not to reassure people that nothing is wrong, but to help distinguish genuine risks from existential hype.

The White House Has an AI Safety Plan, But the Government Lacks the Workforce to Actually Enforce It

atlantic_council

• The White House finalized a framework this summer for testing powerful AI models for safety and cybersecurity risks, but it's voluntary, secretive, and gives the government just 30 days to evaluate each model.
• The federal government faces a fundamental talent problem: roughly 70% of AI PhDs now go into private industry, leaving agencies without the technical expertise needed to independently audit frontier AI systems.
• The authors propose an AI auditing oversight board modeled on the post-Enron accounting watchdog (PCAOB) — a self-funded body that would regulate a market of independent AI auditors, balancing government oversight with private-sector expertise.

In the News

The EU Just Offered Canada 'Associate Membership' — Nobody Knows What That Means, But It Matters

atlantic_council

• European Commission President von der Leyen invited Canada to become the EU's first-ever 'associate member,' but the status doesn't actually exist in EU treaties and the practical details remain almost entirely undefined.
• The offer reflects how Trump's trade war and disengagement from allies is redrawing geopolitical relationships — Canada is now being courted by Europe as a strategic partner precisely because its relationship with Washington has soured.
• For Canada, the move is politically useful even if substantively vague: it signals to Canadians that alternatives to US dependency exist and gives Prime Minister Carney leverage in his ongoing trade standoff with Trump.

Analysis

More Than Half of Women Have Encountered False Claims About Birth Control Online, and Some Are Changing Their Behavior Because of It

kff

• A new KFF survey finds that 53% of reproductive-age women have seen the false claim that hormones in contraceptives are harmful to health, and 43% have encountered the false claim that hormonal birth control causes future infertility.
• One in ten women who saw contraceptive misinformation on social media subsequently stopped or changed their birth control method — a share that rises to 12-13% among women ages 18 to 25.
• About 42% of reproductive-age women have used AI tools like ChatGPT to look for health information, putting AI roughly on par with social media as a health information source, though internet search engines remain dominant at 83%.

Also Worth a Look

• AI Researchers Warning of Human Extinction Don't Have the Evidence to Back It Up (itif)
• AI Is Cutting Off the Entry Ramp for Young Workers — Here's a Ready-Made Fix (democracy_journal)
• AI Could Create a New Kind of Digital Divide — and It May Be Much Harder to Close Than the Last One (brookings)
• A Massive Driver's License Leak Is More Dangerous Than It Looks — AI Lets Adversaries Combine It With Everything Else They've Stolen (csis)
• The US-Saudi Nuclear Deal Could Counter China in the Middle East — But It Comes With Serious Proliferation Risks (csis)
• A 'Super El Niño' Is Coming — and a Warmer World and Global Food Disruptions Make It Far More Dangerous (wri)
• Houthi Advances Are Squeezing Saudi Arabia From Both Ends — and Testing U.S. Credibility in the Gulf (the_conversation)