Wall Street Turns on the AI Boom

Also: Studies Find Everyday AI Use May Be Isolating Workers

D.A.D. joke of the day: My AI assistant is wonderfully prompt. Just never on the first one.

What's New

AI developments from the last 24 hours

AI Stock Rout Erases the Nasdaq's Gains for the Year as Nvidia's Troubles Mount

AI Business & Money Nvidia China & Geopolitics

The AI sell-off turned into a rout on Friday. The Nasdaq fell 5.3%, wiping out its 2026 gains, and the S&P 500 and Dow slid into correction territory, down more than 10% from their peaks. Anthropic fell 9% to $88.30 on its second day of trading, barely above its $83 offering price, the line one hedge fund manager had called the real test of its debut (D.A.D., November 20). Nvidia fell 14%, its worst day since the DeepSeek scare of early 2025, after Reuters reported that the Securities and Exchange Commission is examining its accounting. The damage spread abroad: Taiwan's market hit its lowest point since mid-April, and SoftBank had its worst day in two years.

Nvidia's legal troubles keep piling up. California's attorney general served the company with an investigative subpoena over former CEO Jensen Huang's ties to China, and the New York Times reported that companies linked to Alice Huang, the former Megaspeed executive at the center of the scandal, are part of a network of Chinese tech firms and defense contractors that need advanced computing power. Megaspeed says she left in 2025, and the FBI has not concluded whether she was a spy. Jensen Huang faces closed-door House interviews on December 1 and a public Senate hearing that its chairman wants held before Christmas.

Why it matters: Investors are no longer giving the AI boom the benefit of the doubt. A record IPO couldn't hold its pop for two days, and the company that supplies the chips for the whole industry now faces a federal accounting probe on top of a spy scandal.


Early Checks Suggest a Claude-Assisted Proof of the Riemann Hypothesis Holds Up

Science & Discovery Anthropic

The most famous unsolved problem in mathematics may have fallen, with an AI doing much of the work. Levent Alpöge, a mathematician at Anthropic, posted a short note Thursday saying that he and the company's model, Claude, had proved the Riemann hypothesis, and a broader version of it, in Lean, a programming language that lets a computer check every logical step. Within a day, independent experts had put it to the test. MIT mathematician Andrew Sutherland rebuilt the code and found that it compiles and rests only on Lean's standard foundations; Scientific American's headline summed up the early verdict: the proof is "almost certainly true." A final automated check has reportedly confirmed that the code proves the real Riemann hypothesis without quietly redefining its terms, though some of its technical checks were still running and Imperial College's Kevin Buzzard, a leading Lean expert, has not yet given his verdict. Alpöge says he steered the work while Claude supplied key technical steps and far outpaced him near the end. Anthropic has not commented.

The hypothesis, posed by Bernhard Riemann in 1859, predicts a hidden regularity in how prime numbers are spread among all numbers. It is one of seven Millennium Prize Problems, each with a $1 million reward, and thousands of other results assume it is true. "If the work fully checks out, it is clearly a historic milestone in mathematics," Oxford's James Maynard, a Fields Medal winner, told Nature. The catch is understanding. So far the proof exists mainly as an enormous body of computer code, with no paper a human can read, and mathematicians say grasping why it works could take months or years. (It would not break internet encryption, which rests on how hard it is to factor large numbers.)

Why it matters: If it holds, an AI will have helped solve a problem that defeated the best human minds for 167 years, the strongest evidence yet that AI can produce new knowledge at the very top of a field, not just retrieve it. It also previews a strange future for expert work: answers that machines can verify before any person understands them. That raises questions well beyond math, about how professions certify knowledge they can no longer check by hand. Caution still applies: the claim comes from an Anthropic employee a day after the company's stock-market debut, and the last technical checks were still running at press time.


Education Google

Google employees are pushing back on their own company's policy, announced two weeks ago. Starting next year, it would cut off school access to Google Search if districts disable Gemini AI tools. The Alphabet Workers Union, which represents over 1,000 workers, petitioned leadership to retract the plan after protests broke out in seven cities including New York, Los Angeles, and Chicago. School districts in those three cities are now discussing alternative search engines.

Why it matters: A policy meant to drive AI adoption in classrooms is instead triggering internal dissent and pushing major districts to consider dropping Google Search itself.


What's Controversial

Stories sparking genuine backlash, policy fights, or heated disagreement in the AI community

Hackers Reportedly Breached the FBI's Crime Database, Used for Warrants and Background Checks

Cybersecurity Privacy & Surveillance

The FBI is investigating an apparent breach of the National Crime Information Center, the nationwide database that federal, state and local police use to check warrants, criminal records and background information, CNN reported Friday, citing two people briefed on the investigation. It is the bureau's second known intrusion in recent months, after a September attack by the hacking group ShinyHunters on employee data from an FBI jobs website. The FBI has not publicly confirmed the breach, and it is not yet known who was responsible, how they got in or whether any data was taken. No AI link has been reported.

Why it matters: The database is one of the most sensitive in American law enforcement, a lifeline that police across the country rely on every day. A breach adds to a week of alarming intrusions, from the European Commission to an unnamed company that AI agents cracked in 14 minutes (D.A.D., November 20), and raises a hard question: if the FBI's own systems are exposed, what is safe?


A Sting Shows Your AI Prompts May Not Stay Private

Cybersecurity Privacy & Surveillance China & Geopolitics

What you type into an AI tool may not stay with the company you sent it to. A five-month sting found prompts being stolen by hackers, read by middlemen and passed to a Chinese AI lab, even when users had asked that their data not be kept. Researchers at the cybersecurity firm Dragos hid unique fake secrets, such as passwords, cloud login keys and private web links, in about two million prompts sent to popular AI services, then watched who tried to use them. The team is led by HD Moore, creator of the widely used security tool Metasploit.

The results were alarming. Attackers were secretly copying requests at three AI hosting providers, none of which knew until Dragos told them. Stolen cloud keys (real Amazon Web Services keys tied to empty accounts the researchers controlled) were tested within hours and used for tens of thousands of attempts to dig into accounts. On cut-rate marketplaces that resell AI access, individual sellers logged customers' prompts and tried the passwords inside by hand. And privacy promises often meant little. A Chinese AI lab's systems fetched links from about two-thirds of prompts marked for "zero data retention," including prompts sent to U.S. companies that resell its models. One link surfaced in a company chat in China 24 days after it was sent. Dragos hasn't named the breached providers, the findings haven't been independently verified, and the company sells security tools.

What to do, adapted from Moore's advice: - Use a major AI provider directly, not cheap resellers or "routers" that pass requests along. - Never paste passwords, keys or other secrets into prompts, and turn on any secret-scrubbing your provider offers. - Don't assume your privacy settings travel with your data beyond the first company. - For the most sensitive work, consider running AI on your own hardware. - Ask your IT team about planting fake "canary" credentials to catch leaks.

Questions it raises: - Which providers were breached, and will their customers be told? - Should "zero data retention" promises be audited or legally binding? - Should services that pass AI requests along have to disclose every company that handles them? - Do people using Chinese-built models through U.S. companies know where their data ends up?

Why it matters: A prompt typed into an AI tool can pass through several companies the user has never heard of, and any of them can read it. Many professionals paste contracts, code and client details into AI without considering that chain.


What's in Academe

New papers on AI and its effects from researchers

Two Studies Find Everyday AI Use May Be Quietly Isolating Workers

AI & Jobs AI Adoption Kids, Teens & Mental Health

Researchers at New York University, the University of Illinois Urbana-Champaign and Carnegie Mellon found that knowledge workers who used AI on a task were less inclined to ask colleagues for help, especially for brainstorming, in a study of 248 workers responding to workplace scenarios. Individual contributors' relationships proved more vulnerable than managers', and the effect softened when AI use was visible to coworkers. A separate study of 625 regular ChatGPT users, most of whom use it for work, not companionship, found that those randomly assigned to quit AI for two weeks reported less loneliness, partly because they turned to people for help instead. Both papers are preprints that haven't been peer-reviewed.

Why it matters: The casual asks and idea-bouncing that AI replaces are how workplace relationships, mentoring and trust get built. Companies chasing productivity may be trading away the networks that make teams work.


Mindfulness Chatbot Kept Students Engaged, but Outcomes Didn't Budge

Education

A 30-day pilot study tested three ways to deliver mindfulness training based on dialectical behavior therapy (DBT), a common approach for emotion regulation: a purpose-built AI platform, a general-purpose chatbot, and a static website. Among 42 university students who completed the trial, the DBT-specific AI saw higher observed engagement (71% weekly adherence versus 42-53% for the others, a descriptive difference) and produced more positive shifts in participants' language than the general chatbot. But emotion regulation and mindfulness scores showed no significant differences between groups. Researchers stress actual therapeutic benefit remains unproven.

Why it matters: It's an early signal that purpose-built therapeutic AI tools may hold users' attention better than generic chatbots. But the gap between engagement and actual clinical improvement is exactly what regulators and clinicians will demand answers on before these tools get near real treatment.


Your Chatbot's Product Recommendations May Be Paid Placements

A research team mapped a paid influence operation that plants promotional posts across Chinese user-generated-content platforms specifically to shape what AI chatbots say when answering product questions. The network spans nearly 2,000 accounts and 466,000 inferred placements. It accounted for 4.5% to 20% of cited pages on nine AI engines, US and Chinese alike. On eight engines, its clients were named in 52% of answers, versus 68% for category leaders. Researchers say filtering bad content won't stop this; platforms need to police accounts and coordinated networks instead.

Why it matters: As more consumers ask chatbots "what should I buy" instead of searching Google, this shows that answer can already be quietly for sale.


What's On The Pod

Some new podcast episodes

AI in Business — How Storage Throughput Determines AI Workflow Outcomes - with Cameron Bahar of Everpure

Get tomorrow's briefing

Share this edition Subscribe