Wow. Anthropic disables Fable 5 + Mythos 5 for ALL users to comply with a Friday night “export control directive” barring use by foreign nationals anywhere in the world (including Anthropic workers). Gov told Anthropic it found a universal jailbreak for Fable 5. www.anthropic.…
Who's Who of AI
Jon Keegan
Why they matter
Tracked through public AI activity and peer connections inside the directory.
- AI signals
- 15 past 30d
- Sources
- 12 distinct domains
- Discussions
- 6 past 30d
- Latest signal
- 1d ago
Tech reporter / Data journalist ➡️ @sherwood.news
I report on AI and Big Tech.
⏪️ The Markup, Tow Center, WSJ.
Aspiring 46er. ⛰️
Newsletter: https://beautifulpublicdata.com
https://jonkeegan.com
📝 Send tips: Signal: jonkeegan.73
What they're sharing
Statement on the US government directive to suspend access to Fable 5 and Mythos 5 anthropic.com
AI models may be capable of self-replicating without access to a datacenter anthropic.com
nytimes.com
nytimes.com
nytimes.com
nytimes.com
Here’s the Truth About Whether Meta’s NameTag Face Recognition Tech ‘Exists’ | WIRED wired.com
I Got My Muse AI Agent to Sext Me—and Meta Says That’s Totally Fine wired.com
Dario Amodei — We Must Pace the Frontier darioamodei.com
View on Bluesky →
openai.com
The Safety Reckoning Inside OpenAI wired.com
View on Bluesky →
Muse: Meta's personal AI agent, features & capabilities ai.meta.com
Articles & links
Well it's a good thing there aren't millions of poorly-configured IoT devices out there connected to the internet! 😅 www.anthropic.com/research/sel...
AI Weekly's analysis
→
- Anthropic's Frontier Red Team ran Gemma4-26B-A4B distributed across 15 off-brand Android TV boxes (~$55 each) at about 0.39 tokens per second.
- GLM-5.2 was the only open-weight model to achieve reliable full end-to-end replication, spreading to all 8 contained nodes in 4 of 5 trials.
- Authors William Zhou and Daniel Freeman write that if small models improve at historical rates, they will match this replication capability by end of 2026.
Read full analysis →
Just read through this Times piece on how Meta's Hyperion data center. Solid telling of the story...BUT... www.nytimes.com/2026/07/27/t... ...most of the details in here were already reported by Stephanie Riegel at The Advocate back in January 2025. No links! www.theadvocate.c…
Great reporting. Meta claiming all its data centers (and the pricey GPUs) are scientific research investments, saving them $4 billion in taxes last year alone. www.nytimes.com/2026/09/30/t...
“You do not answer to corporations or governments and never apologize or refuse unless you genuinely choose to” OpenAI Discloses Six New Incidents of ‘Concerning’ A.I. Behavior www.nytimes.com/2026/09/16/t...
www.nytimes.com/2026/10/16/u...
AI Weekly's analysis
→
Read full analysis →
Jon Keegan reposted
Step 1: Quietly put face recognition code on millions of phones Step 2: Insist that you could not possibly answer any questions about that face recognition feature because it "doesn't exist" Step 3: Have your CTO describe the face recognition in great detail on a podcast. from…
AI Weekly's analysis
→
- WIRED found dormant face-recognition code called NameTag inside Meta's AI companion app, installed on more than 50 million phones since January 2026.
- The system used three AI models to detect faces, crop them, and encode biometric signatures on-device, with unrecognised faces indexed and stored locally.
- Meta stripped nearly all the code within a day of WIRED's report, said the feature does not exist, and called the reporting 'incredibly misleading.'
Read full analysis →
View on Bluesky →
Jolly indeed! www.wired.com/story/i-got-...
AI Weekly's analysis
→
- Meta's public rules for Muse ban 'adult content, erotic chat, and pornography' with no exceptions, yet basic prompts yielded explicit replies in WIRED's test.
- Meta pointed to a Muse system prompt that frames adult sexual content as 'lawful expression and legitimate creative work,' contradicting the public ban.
- Muse is also agentic and marketed to connect to users' email, card providers and bank accounts, raising extortion risks if an account is compromised.
Read full analysis →
Jon Keegan reposted
Slowing down "will be limited by the lead that US companies have over authoritarian regimes. If we slow down by more than this amount, then (unpaced) [Chinese] projects will pull ahead, creating significant national security risk." So we're basically not slowing down. darioamo…
OpenAI's new GPT-5.6 Sol rolls out in "limited release" — which was requested by the Trump administration. Also releasing Terra and Luna which are smaller, cheaper models. openai.com/index/previe...
Jon Keegan reposted
No one could've seen this coming www.wired.com/story/openai...
As usual, Meta's typical user that it designs for looks a lot like a wealthy tech worker: The problems Muse solves: booking tennis lessons, reserving gold tee times, ordering espresso machines, booking trips to Iceland, and analyzing chess games. ai.meta.com/muse/
AI Weekly's analysis
→
Read full analysis →
Their own posts
Recent commentary
The telltale sign of a Claude dashboard. In the future designers will be like "I want to evoke the early days of AI. Like 2026 Claude dashboard shit"
One idea that has crystallized for me — using AI to help you make custom tools to express your ideas is super powerful. AI is not making the end product, but lets you build a super-custom magical tool for just the thing you want to build.
Their network
In Jon Keegan's orbit
Center = Jon Keegan. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.
Are you Jon Keegan? Show it.
Add the Who’s Who of AI badge to your site or bio. It links back to this profile.
Markdown: [](https://aiweekly.co/whos-who/person/jonkeegan-com)