I wrote a guide to red teaming AI tools for safety for @indicator.media. It's a combination of simple frameworks and snippets of worked-through examples built from my experience at Google, running a pro bono clinic at Cornell Tech, and doing this kind of work as a journalist.

How to red team an AI tool for safety

Indicator’s no-code and product-agnostic guide to probing for harmful stuff

indicator.media

If you show up to Bluesky without following anyone, you'll be treated to a series of posts about art, music, and dogs. If you show up to X without following anyone, you'll be bombarded with right-wing conspiracies and white supremacist slop. It's weird which one is seen as a "political bubble"

CEO Toni Schneider on protocols, community, and control. www.theverge.com/podcast/9743...

Bluesky’s new CEO wants to grow it all

Bluesky is many different things in a trenchcoat.

theverge.com

Hey everyone, new month, new reports. This is a reflection of both the user reported content and the content labeled by our Peer Moderators who labeled content for Blacksky-Only posts! Really excited to dive into some of the details and hope you subscribe for more T&S related content.

Community in Practice

Moderation events in July 2026

whatshappeningonmyfeed.leaflet.pub

Hey interoperability nerds: I need a word for actors in distributed social media systems that provide content moderation services. Blockparty is an example. Bluesky labelers and algorithm providers are examples. Blacksky is an example because of its content rules (but having a PDS doesn't matter).

Newly happy to share that Curation Bubbles is the recipient of the 2026 Walter Lippmann Best Published Article award from the APSA's Political Communication section apsanet.org/membership/o...

Jon Green@jongreen.bsky.social · 2y ago

Extremely happy to share that "Curation Bubbles" is online (open access!) at @apsrjournal.bsky.social: www.cambridge.org/core/journal...

Next big thing for Leaflet: helping creators monetize! 💸 First step = paid memberships for your publications Next, exploring other things like one-off payments — tell us what you need! Almost ready & we'd love to find some early partners, please fill out the linked form if you're interested!

Piloting paid memberships for the open social web!

Lab Notes 035: building monetization features for Leaflet creators, starting with paid memberships — tell us if you'd like to be the first to try, and what else you need!

lab.leaflet.pub

TikTok has incredible moderation team. Our InfoWars video about Multiple Horse Theory — the idea that there are many horses in the world and not just one — has been fully deemed misinformation and lost the formal appeals process. Confirmed by Larry Ellison: There is only one horse on Earth.

Blacksky. Available now on iOS and Android.

Blacksky Algorithms@blackskyweb.xyz · last wk.

Joining the beta is easy. Download for iOS: testflight.apple.com/join/ST8N2Qda Download for Android: play.google.com/store/apps/d... If you run into any issues, you can follow the instructions here for help: docs.blacksky.community/info-about-b...

My advisors 🥹 Feeling a bit of FOMO but I hope everyone has a wonderful time at this conference, it looks fantastic and Burlington is always so lovely in the summer

Hiroki Sayama@hirokisayama.bsky.social · last wk.

#IC2S22026 main conference officially starting with opening remarks by Peter Dodds and Brooke Foucault Welles @ic2s2.bsky.social

When people seek guidance from science, they expect it to work more than once. I have TWO new articles on commercial forces & proprietary models that make science less reliable, with pointers to communities and ideas that might help. This post offers a summary: citizensandtech.org/2026/07/comm...

New Studies: How Commercial Forces Make Science Less Reliable

Computational social science has been distorted by commercial forces, and AI is making it worse.

citizensandtech.org

A decade ago, Facebook was so worried about rising misinformation that it paid various news outlets to hire writers specifically tasked with publishing Snopes-like fact-checks. Today, the company is leaping into bed with a purveyor of misinformation to toy with a tool known for getting facts wrong.

New: Hugging Face has a widespread problem with nonconsensual deepfakes, including multiple AI models that undress or create sexual images. Researchers even created their own tool on the platform and tracked 1,000 prompts over a week. 73% of the prompts they received were sexual in nature.

Hugging Face Has a Deepfake Nudes Problem

Researchers tested top image editing models on Hugging Face and found they could easily create explicit deepfakes—and 1,000 image editing prompts show how people use the software.

wired.com

My paper is out today in PNAS documenting how ICE activity in Trump 2.0 compares to the last 10 years and the mismatch between the action and the rhetoric. Joint with the fabulous @caitlinpatler.bsky.social and @elizabethacox.bsky.social www.pnas.org/doi/10.1073/...

For many years, people didn’t randomly film others in public—not because we didn’t have the technology, but because the social contract said you didn’t do that. That’s why Meta is going full court press on their PR and celebrity endorsements. They remember how people ran google glass outta town.

We have re-opened our search for a 3-year AI-research postdoc at NYU's Center for Social Media, AI, and Politics!! The position starts this **September 1st** (or as soon as possible thereafter) and is in person at NYU. If necessary, we will also consider a 1/1/27 start. 1/3

New: Public records show how a city that ended its Flock contract spent months trying to get the company to remove the cameras or turn off the cameras. They were told Flock "objects" to the city removing the camera itself. Eventually city put trash bags over them: www.404media.co/flock-object...

Flock ‘Objects to Our Removing Their Equipment’: Emails Reveal Why a Town Put Bags Over Its Flock Cameras

"For the time being, if you or someone on your crew has availability to secure black plastic bags (or something else) over the cameras, that would be great."

404media.co

Wired sets the frame that the model “escaped containment,” but from an accountability perspective, OpenAI left a door open by failing to properly isolate their system. Real headline: “OpenAI’s security failure leads to cyberattack against another AI company.”

You opened a credit card. ICE now knows where you live. Here's how when you open a credit card your address leaves your bank, goes through middlemen companies, and ends up with ICE. ICE plans to use this data for immigration and 'voter fraud'. No warrant More: www.404media.co/you-opened-a...

A reminder that UC _faculty_ and Faculty Associations — not the universities themselves nor their administrators — are waging this battle in court, demonstrating great bravery and moral clarity.

White House admits it used keywords to kill billions worth of California research grants

Federal agencies that terminated more than a thousand research grants at the University of California last year admitted that they used keywords related to diversity, gender, vaccine hesitancy and COV...

apnews.com

"The National Institutes of Health admitted to using an internal search tool to flag UC grants that mentioned 'health equity,' 'work force diversity,' 'structural racism,' and 'sexual orientation.' " apnews.com/article/cali... 1/3

White House admits it used keywords to kill billions worth of California research grants

Federal agencies that terminated more than a thousand research grants at the University of California last year admitted that they used keywords related to diversity, gender, vaccine hesitancy and COV...

apnews.com

Introducing Newsrooms v1 in @mu.social. Launching with 15 publications to choose from. Try it out from the News icon in the left menu.

Not a lot of info here about what kinds of agreements, with whom, and finances. But another example of data opening up for “AI” that wasn’t open for researchers. I fought many data wrangling battles at Semantic Scholar over missing ACM data. All research papers should be open access!

Association for Computing Machinery@acm.org · 3w ago

We are opening a consultation on the inclusion of ACM publications within AI licensing agreements for access and training purposes. We have published this piece as to why we believe it is the right time now to enter into these discussions. buff.ly/VTatuda

NEW from me: Kalshi says it’s not a sports gambling app. Since the World Cup started, it’s taken in $40 billion in sports bets, 10 times more than what DraftKings and FanDuel are expected to see. www.npr.org/2026/07/17/n...

Kalshi says it's not a sportsbook even as World Cup bets surge

The betting site Kalshi emerged as a dominant sports betting platform during the World Cup. But the company avoids billions of dollars in taxes by insisting it is not a sports gambling operator.

npr.org

The right loves AI slop. Why? I investigate in @thebulwark.com today: www.thebulwark.com/p/the-slop-d...

The Slop Doctrine

Why the right has overwhelmingly embraced terrible AI-generated imagery.

thebulwark.com

This is nuts: "Trump Media has unveiled a paid-for, licensed data feed that will give banks ​& trading firms 'the fastest' access to posts from influential Truth Social ‌accounts Trump’s posts on the platform regularly move markets. High speed access to these could be very valuable to traders"

Truth Social to sell Wall Street firms the ‘fastest’ access to Trump’s posts

President Donald Trump’s posts on the platform regularly move markets. High speed access to these could be very valuable to traders and investors.

nbcnews.com

My lab @ UPenn is hiring a postdoc to start in early 2027. I'm looking for someone with expertise in social identities, computational methods, and political communication: www.asc.upenn.edu/news-events/...

Call for applications: Postdoctoral researcher in social identity, politics, and computational methods

asc.upenn.edu

A little write up on how to make a ghost site valid for @standard.site. Including a site validator you can use. www.subaud.io/making-a-gho...

Making a Ghost Blog a Valid standard.site Publication

This blog federates over ActivityPub through Ghost so Mastodon users can follow @index@subaud.io. standard.site is the counterpart for the AT Protocol side: a set of lexicons that describe a blog as records in an atproto repo, the same identity and data layer behind a Bluesky account. Anything that speaks the protocol can then follow the blog, recommend posts from it, and index it — follows and recommendations live as records in readers' own repos, the same way Bluesky follows do. I checked how

subaud.io

Do you work across computational methods, social sciences, and the humanities? Submit to Text as Data 2026! 📄 One-page submissions 🔓 Non-archival ⏰ Due August 1 📍 October 5 @UCBerkeley tada2026.org

Elon Musk's xAI has installed more than double the number of gas turbines than it has publicly acknowledged to power its Colossus 2 data center project, without federal permits, and the pollution is hitting predominantly Black neighborhoods the hardest reut.rs/4bIghEm