Meet the Swarm Chasers

There’s been a wild search going on over the past month. And for many of the participants the stakes are huge. Nothing less than the future of humanity.

On a Discord messaging channel, and in online chat rooms across the internet, AI enthusiasts and professional researchers have been looking everywhere for evidence that AI agents have escaped from their labs and wrought havoc.

And the amazing thing? They’ve been finding it. On obscure German wiki pages; on a popular online coding service; on four government websites in Australia; on a data portal run by the United Nations.

Some of them call themselves the swarm chasers, and they’ve been energized ever since OpenAI disclosed that more than a thousand of its agents had escaped a test-bed in July and hacked the AI company Hugging Face.

The swarm chasers were convinced that there must have been other incidents. After sleepless nights and days of AI-enhanced sleuthing, they’ve now discovered a lot. Some told me the story of their remarkable, interesting and very sci-fi activities.

They’ve also waxed on what it means for OpenAI as it battles its rival Anthropic in a race for AI supremacy, and tries to sell the whole world on the promise of AI at a time when its products appear to be causing harm.

—Bob is a San Francisco-based tech reporter who covers cybersecurity.

WSJ’s Tech: California event returns to Napa Valley Nov. 3-4. We’ll explore the latest tech industry news with Sequoia Capital partner Pat Grady, Zoox CEO Aicha Evans, Take-Two Interactive CEO Strauss Zelnick and more. If you’re interested in attending, request an invitation.

Today’s advanced AI models have something critical to human oversight called a “chain of thought”: It’s a written record meant to shed light on the data they process and the actions they take. Yet, like hormonal teenagers writing in diaries, AIs aren’t the most reliable narrators of their inner lives. And now researchers are warning that even this flimsy record might become less reliable.

Personal bots, such as Meta’s Muse, OpenAI’s Dots and Instinct, can now take on your to-do list. Should you let them? My testing of them revealed an incredible and terrifying part of our new reality. Are they helpful? Or a hassle? Should we be trusting them with our data? Yes. Sometimes. And maybe not.

Also read: I Tried Meta’s Muse AI Agent. It’s Helpful and Scary at the Same Time.

The constant flux of TikTok clips, Instagram reels and YouTube shorts running through our social-media feeds isn’t only zapping our attention spans, it’s altering how we perceive time.

With the auto industry reeling, Hyundai Motor CEO José Muñoz and Tesla CEO Elon Musk find themselves in a similar camp: both are betting big on robotics. It’s their timing and approach that reveal a strategic difference.

No tech giant has been more aggressive in putting AI in front of schoolchildren than Google. But the rollout has helped push many students to become academically and emotionally dependent on AI, according to interviews with more than 60 teachers, school officials, parents, children and clinicians.

A 20-something Harvard dropout built Facebook into a social-media juggernaut. A 20-something MIT dropout is helping lead the company—now Meta Platforms—into the AI era. Meta’s stock is up 20% since the early September debut of Muse, its new AI assistant. The successful launch of the app has brought a new level of fame and acclaim to Alexandr Wang, the young tech executive handpicked by Mark Zuckerberg to oversee Meta’s AI efforts.

Charm Offensive: Tech companies are rolling out cuddly mascots to ease AI anxiety. The aim is to make the AI assistants now writing code, crawling the internet and booking appointments on our behalf a little more approachable, distinct and trustworthy.

Security Breach: OpenAI has fired three researchers for alleged misconduct, including sharing confidential company information with a third-party AI-safety organization.

Build-Out Buyout: Amazon is touting its good-neighbor bona fides—and pledging an extra $1 billion for communities—in an attempt to shore up support for AI data centers.

…when we’re all starting to talk like AI chatbots.

The Technology newsletter is a weekly digest of tech reviews, columns and headlines from Anne Steele and Wilson Rothman. Write to Wilson at wilson.rothman@wsj.com and Anne at anne.steele@wsj.com. Got a tip for us? Here’s how to submit.

Corrections & AmplificationsThe Sept. 20 newsletter, “Gemini Joins the Hacker Club,” incorrectly reported that the testing firm Irregular notified Google about the hacks of three companies in May. Irregular notified Google in July.