Hey-o, it’s Monday! This weekend, while you were contemplating the eerie, not-quite-human qualities of a supposedly intelligent entity known as former Theranos CEO Elizabeth Holmes, AI news chugged right along.

In a throwback to 2023, Meta alarmed people with a short-lived plan to gobble up mountains of customer data for AI-training purposes; we learned about a study that revealed what you probably already knew about AI’s effects on your memory; and we found out Apple’s AI-related delays kept its latest gadgets boxed up in warehouses for a hilariously long time.

Let’s get started.

→ Meta’s training data overreach
→ Researchers find AI-altered images can be used to incept you
→ AI caused Apple’s latest gadget to collect dust in warehouses for almost a year

The news: A piece of analysis from Nieman Lab originally released on Thursday set in motion a backlash that Meta was dealing with well into the evening on Friday. As Tom McKay wrote for Gizmodo, the move looked like Meta was trying to “shore up the company’s legal footing as it stakes its future on web-browsing agents like its Muse AI agent.” Muse, Meta’s agentic AI platform pitched at normies rather than coders, has attracted surprisingly positive mainstream press coverage, and injected some swagger into CEO Mark Zuckerberg and AI Chief Alexandr Wang—at least if you judge from their beer-guzzling podcast victory lap.

The extraordinary terms, which weren’t set to go into effect until Oct. 30, would have given Meta permission to use business websites’ data to “provide, develop, improve, research, train, and maintain” its own products. But on Friday evening, Meta backtracked, at least slightly.

The takeaway: Meta’s Friday statement says, in part, “We’ve heard you and want to clarify: the crawling terms we consolidated only applied to our ads and business products, and we’ve clarified that in the terms themselves.” Crucially, the word “train” was removed from the terms.

The news: Your memory—yes, even yours—is full of lies. Maybe you already know this, because you’ve already read about how even people with powerful recall abilities can have their memories manipulated. But as fun as it sounds to mess up people’s recollection of the past, I know what you’re wondering: can I speed the process along with AI? And the good news for all you gaslighters and con artists is, yes you can, new research has found.

Participants in an MIT study were shown photographs, then after a break, they were shown either the same photographs, or AI-altered versions of them. In a memory test afterward, people shown the AI-altered versions remembered the original photos incorrectly, and displayed totally unearned confidence in their false memories.

The takeaway: This isn’t the first study on AI implanting false memories, and it probably won’t be the last. As AI makes inroads into our daily lives, we’re either going to have to learn to counteract its negative effects on cognition, or collectively let all our brains turn to mush.

The news: The intended release timeline for Apple’s HomeView did not involve the device—with its 2024-era processor and it’s 8 gigs of memory—being released in late 2026. According to Bloomberg’s Mark Gurman, the devices have been boxed up and ready in warehouses for “the better part of a year” while they waited for Apple to ship Siri AI, a major component of the software that powers them.

This may seem like ancient history now, but for years, Apple struggled to implement AI. Apple had to pull a TV ad after the company failed to make the AI features shown in the commercial a reality, and faced legal problems over alleged broken AI promises to consumers. Apple finally threw in the towel and essentially rented Siri a brain from Google in the form of an AI model. But the result of that surrender is a version of Siri that puts user-friendly AI features on iPhones, and enables Apple to, at long last, release a whole new line of products, namely the HomeView.

The takeaway: You can get your AI-powered HomeView on Nov. 13, but you might have to blow some dust off the box before you open it.

Sanders has introduced the Ban Artificial Superintelligence Act with Rep. Greg Casar (D-Texas), legislation that would establish a federal cabinet-level Department of Artificial Intelligence.

Read more →

The Wall Street journal reports on why AI is becoming a nightmare for the insurance business.

Disney hoovers up employees from the smoldering remains of Character.AI, gives them huge pay packages.

Could Alexandr Wang be Mark Zuckerberg’s Padawan?

DeepSeek, known for fast AI models, releases an even faster model.

Anthropic is what’s known as a public benefit corporation. It just announced a “Public Benefit Fund” to—in theory—help realize all that public benefit.

Bernie Sanders argues that humanity must not allow "a handful of billionaires to recklessly replace us."

OnlyFans creators are frustrated that Meta lets Muse, its AI agent, engage in X-rated content but limits everyone else.

Data centers are taking their messages to schools with activities and holidays like "International Data Center Day."

The activist behind the world's first 3D-printed gun is claiming another first: "the first AI-made gun".

Young AI cloud providers, known as neoclouds, are a dime a dozen right now. Experts say only seven to 10 will remain in the future.

At least 47 large data center projects have been delayed in 2026, casting doubt on the feasibility of America's AI infrastructure ambitions.