Meta Now Lets Anyone Use Your Instagram Photos in AI Images—Unless You Opt Out
Meta’s new Muse Image model — built by its Superintelligence Labs division and rolling out across Instagram, WhatsApp, and the Meta AI app — can pull real Instagram users into AI-generated photos, and the default setting is opt-in on your behalf. If you have a public account, congratulations, you’re a training asset and a generative subject unless you go find the buried toggle to say otherwise. The gap between “we gave users control” and “we gave users the illusion of control” has never been narrower.
OpenAI’s Chief Futurist Is Leaving the Company
Joshua Achiam — nearly nine-year OpenAI veteran, AI safety researcher, and someone most people only learned existed during the Musk v. Altman trial — is heading for the exit. The departure of a long-tenured safety-focused researcher at a company that just restructured into a for-profit and is sprinting toward AGI is the kind of thing you file under “probably fine, definitely worth watching.” The revolving door on OpenAI’s safety wing keeps spinning.
Anthropic is launching Claude Cowork on mobile and web
Claude Cowork — Anthropic’s agentic platform that handles long-running tasks autonomously — is finally leaving the desktop and coming to mobile and web, starting with Max subscribers. The pitch is genuinely compelling: kick off a task at your desk, close your laptop, and check in from your phone when it’s done. Whether “your AI coworker never stops working even when you do” is reassuring or existentially exhausting probably depends on your relationship with your own inbox.
Discord admits AI moderation bug wrongfully banned users over harmless images
Discord’s automated moderation system has been incorrectly banning users since May — apparently flagging harmless images as violations — and it took until this weekend before the team identified and fixed the problem. At least 200 additional users were banned over the weekend alone before anyone noticed. This is the core tension of AI moderation at scale: it’s fast, it’s cheap, and when it’s wrong, real people lose access to their communities with no one home to answer the phone.
Microsoft joins AI cost-cutting trend by relying more on its own models
Microsoft — which wrote some of the largest checks in AI history to OpenAI — is now quietly shifting workloads to its own in-house models to cut costs. There’s something almost poetic about the company that bankrolled the AI arms race now looking for the discount aisle. It also signals that the “just use the frontier model for everything” strategy has a ceiling, and every major player is starting to hit it.
Cops Say Waymo Snitched on Teens for Allegedly Drinking and Shooting a Toy Gun
Waymo called San Mateo police on teenagers inside one of its vehicles who were allegedly drinking and firing a toy gun, prompting the department to post on Facebook: “Parents do you know where your teens are? Waymo does!” Setting aside whether that’s the right call, we’ve now fully arrived at the era where your ride-share vehicle is also a surveillance node with agency — and the cops are cheerfully marketing it. The smart glasses surveillance debate just got a four-wheeled competitor.
Avoid AI atrophy — new tool promises to reverse vibe coding skills decay
A new CLI tool called Atrophy is pitching itself as the solution to developers losing their actual coding chops by over-relying on AI assistants — the idea being that if you don’t use your skills, you lose them. It’s a genuinely interesting counter-move in the vibe coding moment: the same week Godot banned AI-generated contributions, we’re seeing tooling emerge to remind developers that there’s still a brain behind the keyboard. Whether this catches on or becomes the coding equivalent of a gym membership people buy in January remains to be seen.
Erling Haaland Is Everywhere at the World Cup. Most of It Is AI
Norway’s Erling Haaland has become one of the World Cup’s most ubiquitous figures — in AI-generated memes, fake highlight packages, and synthetic content flooding social feeds — despite just being a footballer who showed up and played. It’s a useful case study in how generative AI is reshaping celebrity and fandom: the line between the real person, their curated image, and the character the internet builds around them is now officially blurry enough that most people won’t bother to look for it.
Bottom Line
The week’s pattern is crystallizing: AI is getting faster, cheaper, more capable, and more embedded in everything — and the trust, safety, and consent questions are running about three sprints behind.