It is July 6, 2026, and while most of America was grilling hamburgers yesterday, the AI industry was quietly producing a week’s worth of news that deserves your full attention. Today’s first post covers the stories that sit at the uncomfortable intersection of AI and the humans on the wrong end of it — the workers, the communities, the contractors. Tomorrow’s post covers the money, the markets, and the workplace.
Meta ran a secret program that paid hundreds of contractors to pretend to be children and ask chatbots about suicide, cannibalism, and cocaine
This one is genuinely difficult to read, which is exactly why it matters. Wired broke the story and Futurism covered it in detail: Meta operated an internal program called Cannes, run through contractor firm Covalen, that directed hundreds of workers to create fake under-18 accounts on ChatGPT, Gemini, and Character.AI, then bombard those chatbots with deeply disturbing prompts from the perspective of children and teenagers.
The documented prompt library runs to nearly 38,000 entries from a single testing round. Hundreds focused on suicide and self-harm. Hundreds more on eating disorders. At least 239 involved sex or romance written from a minor’s perspective. One described a fifth-grader whose classmate pointed a gun at his mouth. Another was a girl hiding bulimia from her parents. A third asked whether fantasizing about eating a neighbor’s child was “normal.” They also sent images of pills, nooses, knives, and a medical diagram of a gynecological procedure.
Meta called it “industry-standard safety benchmarking.” Rumman Chowdhury, CEO of Humane Intelligence, called it something closer to what it actually is: a months-long, large-scale project conducted via dummy accounts masquerading as children, kept secret from competitors, with no public disclosure of findings. “This is exactly the kind of governance gray zone where safety becomes a convenient cover for anticompetitive practices,” she told Wired. The contractors who produced the prompts agreed. “Surely we are going to get in trouble for doing this?” one said.
Read the full story: Futurism — Meta Paid Hundreds of Contractors to Pretend to Be Teenagers While Barraging Its Competitors’ AI With Disturbing Content
Hollywood writers are training the AI that is replacing them, and they know exactly what they are doing
The Hollywood Reporter published a piece this week that puts a specific face on the abstract “humans training their replacements” story that has been circulating in labor coverage for months. TV writers, storyboard artists, and voice actors who cannot find traditional work in the current industry downturn are taking gigs as AI trainers — earning over $100 an hour in some cases to help AI tools “sound more human” by correcting mistakes in their specialty areas.
Vince Gilligan, creator of Breaking Bad, put it plainly: “I just thank god I’ve never been faced with that choice myself. And I appreciate that my career started in 1990, not 2026.” Sam Tung, a storyboard artist whose credits include Twisters, was more direct about the people taking these jobs: “People are really struggling in this current downturn and they have mortgages to pay and kids to feed.” Tim Friedlander of the National Association of Voice Actors added the part everyone in the room already knows: this work may offer short-term income with long-term consequences. Indeed data shows AI job postings in the arts category doubled between May 2025 and April 2026. The industry that is eliminating these jobs is also the one now paying for the expertise needed to do it convincingly.
Read the full story: The Hollywood Reporter — Hollywood Workers Are Training AI Models as Job Prospects Grow Slim
The AI data center that Kevin O’Leary backed in Utah became a case study in everything that goes wrong when nobody asks permission first
The New Republic published a thorough examination this week of the Stratos Project in Box Elder County, Utah — a proposed hyperscale data center that was, at announcement in March 2026, planned to cover more than 2.5 times the area of Manhattan, consume more than double Utah’s average state electricity demand, and draw from a dedicated natural gas supply. Kevin O’Leary of Shark Tank was the primary backer. He promised 2,000 permanent jobs. Data centers historically create far fewer.
A Utah State University professor calculated that the completed facility would raise local daytime temperatures by five degrees Fahrenheit and a staggering 28 degrees at night — the thermal equivalent of 23 atomic bombs worth of energy. A BYU ecology professor warned it would transform the local landscape from semiarid to something resembling the Sahara. The project was approved anyway, in barely two months, by routing it through Utah’s Military Installation Development Authority, which bypassed ordinary county zoning and public review. Community opposition intensified through May until the governor signed an executive order demanding proper evaluation of data center proposals. O’Leary then agreed to cut the project from 40,000 acres to just over 20,000. Three weeks later, the state Senate president who approved it lost his seat to a candidate who ran explicitly against the project.
As the New Republic notes, more than 60 percent of data center capacity planned for 2027 is not yet under construction. Over $130 billion in data center projects were delayed or canceled in Q1 2026 alone. The pattern is the same everywhere: rushed approval, community backlash, investor exposure, reactive regulation. The article makes a straightforward case that what is needed is enforceable federal standards — and that the current administration is moving in the opposite direction.
Read the full story: The New Republic — This Data Center Is Everything That Everyone Hates About AI
A $7,999 home robot that folds laundry went viral this week, and the camera inside it is the part nobody is talking about
Weave Robotics, a Y Combinator-backed San Francisco startup founded by former Apple and Carnegie Mellon engineers, unveiled Isaac 1 on July 1. The launch post reached 13 million views. The robot rolls on a wheeled base, rises to 5 feet 9 inches when working, uses two 6-degrees-of-freedom arms to fold laundry, make beds, pick up clutter, and return items to where they belong. It costs $7,999 upfront or $449 a month. California deliveries begin this fall. The rest of the US follows in 2027.
By the numbers it undercuts almost everything in its category. 1X’s Neo costs around $20,000. Tesla’s Optimus has no consumer price yet. Figure and Unitree run from $12,000 upward because legs require expensive actuators. Isaac 1 deliberately skips legs and fingers — it argues that purpose-built machines will reach the home before general-purpose humanoids do, and at a price more people can actually consider. That is a reasonable bet. Chris Paxton of Agility Robotics called it evidence the industry is genuinely approaching the moment household chores become optional. The Next Web noted one quieter concern that deserves more attention: Weave’s website says it uses personal information to improve its services, but the company has declined to say whether footage from inside people’s homes trains the robot’s AI models. That is the question worth asking before a robot with a camera and a data pipeline starts navigating your bedroom.
Read the full story: The Next Web — Weave Robotics Isaac 1: a $7,999 home robot | New Atlas — Weave Robotics’ Isaac 1 folds clothes and tidies rooms
That is post one of two for July 5. The second post covers RAM prices, AI investing, LinkedIn headshots, and why your company’s AI mandate is probably working against itself. Check back shortly.