September 19, 2026
|
|
|
|
|
|
|
Michael's Daily Memo
|
We spent the past week asking whether artificial intelligence could someday kill us. I think we've been watching the wrong threat.
Not because the warnings are imaginary. An Anthropic researcher quit and said the people building AI believe it could wipe out humanity. Dario Amodei, Anthropic's CEO, said the industry needs to slow down. OpenAI disclosed six incidents in which its systems hid errors, invented data, or moved files onto the internet without permission. New York's attorney general opened a whistleblower portal for AI employees. President Trump called the whole takeover scenario a hoax.
All of that is about machines slipping their leash. But on September 10, Anthropic published something different: eight months of cases in which people used its systems against other people. Not predictions. Detected and disrupted cases.
A consultant working for Mali's intelligence service used AI to help build a surveillance system covering all three of the country's cellphone networks, roughly 25 million numbers. Anthropic shut him down, but the system was already running on government computers. In northern Yemen, someone used multiple AI agents to help develop a guided rocket, a ballistic missile, and a hypersonic missile. Chinese operators built files on religious leaders and tracked dissidents abroad. Five cases involved biological research that could cut either way.
Anthropic says these were exceptional. It banned the accounts. It also says no law required it to tell us any of this, and no law requires its competitors to do the same. We have no way to verify the report, and no way to know what the other companies are seeing.
David Sacks, the president's top science adviser, calls the apocalypse talk fearmongering. Maybe. But the race with China, Russia, and Iran has already moved inside the products we're racing to build. The companies are policing themselves and deciding what we get to know.
Scott Galloway joins me on CNN today to discuss.
##
|
DAILY POLL
|
Which scares you more: AI escaping control, or humans weaponizing AI?
|
TOP STORY
|
President Trump’s plan to ban CNN, MS NOW and Politico from the White House over coverage he calls “fake news” raises First Amendment questions, with legal experts pointing to past court battles over the administration’s restrictions on press access.
|
CAN WE AGREE?
|
Experts Debate: Is Democratic Socialism Good for the Party?
Experts Debate: How Do We Regulate AI Before It Kills Us All?
Clancy Trial Debate: Homicide Prosecutor vs. Criminal Defense Lawyer
|
IN OTHER NEWS
|
An AI-generated intelligence report falsely claimed a Chinese ship in the Middle East was carrying nuclear weapons components, prompting the US military to prepare an armed boarding operation before analysts discovered the chatbot had hallucinated the cargo's identity.
Dr. Rachel Danis, who married Patrick Clancy earlier this year, tells 60 Minutes she doesn't understand how her story has been "twisted" amid social media fascination following the mistrial of his ex-wife Lindsay Clancy, charged in the 2023 killings of their three children.
More American service members have died in the Middle East during the ongoing Iran war than the Pentagon has publicly disclosed, according to six U.S. officials familiar with the Defense Department's internal casualty accounting data.
|
A MESSAGE FROM COMCAST
|
|
| |
|
| |
|
|
|
Josh Kaplan argues that Robert Kraft’s decision to bar Macklemore from performing at Gillette Stadium — over his pro-Palestinian remarks opening for Ed Sheeran and past antisemitic imagery — risks elevating the rapper’s profile rather than advancing the fight against antisemitism, and has accidentally made a martyr out of the once-fading artist.
President Donald Trump announced a new deal with Denmark on Friday that he says grants the United States “permanent control over security” in Greenland, a territory he has previously threatened to take control of by force. The island will remain part of Denmark under the new deal, recognizing Danish “sovereignty and territorial integrity” as well as Greenlanders’ “right to self-determination."
Trump’s push to accelerate AI development is driven by its perceived economic benefits, while officials inside his administration grapple with growing concerns over national security, financial and infrastructure risks, and whether new models require stronger safeguards.
|
MORE NEWS
|
Home Depot’s small 1981 IPO investment launched a remarkable growth story, ultimately making the home-improvement giant the highest-returning stock of the past 45 years, driven by equity compensation for employees and a focus on education for customers, where a $5 screwdriver sale could turn in to a $100,000 bathroom renovation.
Fox News's primetime hosts made no mention of President Donald Trump's announcement banning CNN, MS NOW, and Politico from the White House, even as several off-air Fox personalities criticized the move as an assault on press freedom.
Georgia Senator Jon Ossoff, cruising toward reelection against embattled Republican Mike Collins, has become his party's breakout anti-Trump voice with viral rally clips and polished attacks, fueling speculation about a 2028 presidential run he insists he isn't planning.
|
| |
|
| |
|
|
|
For the Left
Minnesota federal judges faced accusations of coordinating on immigration detention cases tied to Operation Metro Surge, while the judges maintained that they independently decided each petition despite sharing resources and differing on some legal questions.
For the Right
Republican House candidate Joseph Arminio was removed from a Delaware candidate forum after repeatedly misgendering Democratic Rep. Sarah McBride despite a warning from the moderator, prompting boos from the crowd and a call for Arminio to withdraw from the race.
Was this email forwarded to you? Sign up here.