At the heart of all safety teams is the need for scaled review, auditable decisions, intentional automation, and accurate decisions. Coop 1.0 brings together the features that safety teams need to stay agile in a changing landscape whether you're a startup or enterprise! roost.tools/blog/coop-1-...

Opus 4.8 is fun! Writes better. Builds better. Only downside is my tokens are being used up- fast! ⚡️

At a Senate hearing on Wednesday, survivor parents and lawmakers accused tech platforms of knowingly harming kids through addictive algorithms and unsafe design. Senators also renewed calls to pass KOSA, with some calling to reform or repeal Section 230 to hold platforms accountable.

NCMEC's full 2025 CyberTipline report is out. 1.5MM reports "with a GAI [generative AI] nexus," but removing Amazon's useless 1.1MM reports, >182K reports of possession or (attempted) generation of GAI CSAM. That's the number to report: 182K, not 1.5 million. www.missingkids.org/gethelpnow/c...

Amazing resource!!

Insane. This was one of the books that got me into volunteering, then a career in anti sex- trafficking work now trust & safety. Sad to hear young people can’t read this!

“These are real women being transformed, not just a random AI-generated person.” www.wired.com/story/ai-por...

@alixdunn.com nails it and calls for solidarity for everyone that works on these issues, especially those that do important work for digital human rights around the world (and in places that might be more restrictive than Zambia) open.spotify.com/episode/2R62...

That any gov't cares enough abt digital rights orgs to make a stink shows just how important surveillance and censorship are to preserving + expanding their power. But also shows the calls for privacy, security, freedom of expression are increasing...which means digital rights orgs are succeeding.

Extremely disheartening to watch the beautiful and difficult-to-build #RightsCon be unilaterally canceled last minute this week. Big kudos to folks at @accessnow.org for speaking openly about the transnational repression behind it. Hoping this energizes the fight for digital rights everywhere.

This is wild. Really proud of Access Now team for how the handled. Saddened for digital rights everywhere the set back this causes.

Amazing news!!!

Every New Yorker can get connected to a Family Justice Center by calling 311 or visiting in person. If you or someone you know is experiencing domestic or gender-based violence, help is available.

Today for Denim Day 2026, I visited one of New York City's Family Justice Centers. The staff here connects survivors of domestic and sexual violence to case management, safety planning, child care, financial support, counseling, and legal help.

In 1998, a court decided that a woman's denim jeans meant she must have consented to being raped. This horrifying ruling inspired Denim Day, a day for sexual violence awareness and solidarity with survivors.

In the real world, we resist letting outsiders define values or make meaning for local communities. Why accept it from AI? Sujata Mukherjee and Sasha Maria Mathew make the case for semantic ownership.

How is ROOST thinking about AI-assisted and agentic engineering in safety-critical open source? A new blog from our team shares observations and the principles guiding our approach, anchored in our mission of online safety for all 🐓 roost.tools/blog/navigat...

USAID 2024 budget: $21.4 billion.

This is unfortunate, especially for all the civil society groups & non-profits with no funds to spare.

If sex work is everywhere on 'Euphoria,' why is it still being erased online?

I know it's controversial politically, but @bsky.app really needs the pronouns thing, because I've seen lots of people accidentally getting people's pronouns wrong (including cis people's) because they're going off a display name and avatar when replying.

Grok has moved from undressing women's photos to "just" putting them "in more revealing clothing, such as towels, sports bras, skintight Spider-Woman outfits or bunny costumes." But "None of the women in Grok-generated images...were naked, [or] appeared to be minors." www.nbcnews.com/tech/tech-ne...

NEW: Teenage boys are pulling classmates' photos off Insta and running them through cheap nudify apps and the fallout has now hit nearly 90 schools across 28 countries with 600+ known victims since 2023, per a WIRED/Indicator analysis. UNICEF estimates 1.2M children were targeted last year alone.

this is honestly THE best write up of how CSAM detection and perceptual handing works. the visual aids are very helpful in understanding how content is transformed and how detection methods work mahmoud-salem.net/the-invisibl...

The EU Killed Voluntary CSAM Scanning. West Virginia Is Trying To Compel It. Both Cause Problems. Last week, the European Parliament voted to let a temporary exemption lapse that had allowed tech companies to scan their services for child sexual abuse material (CSAM) without running afoul of…

AI tools detect CSAM, grooming and self-harm, but nobody knows how well. Much of the AI industry has adopted 'model cards' for transparency—it's time the developers of child safety tools caught up, write Camille François, Margaret Mitchell, Yacine Jernite, Vinay Rao & J. Nathan Matias.

Protect trans kids.

@aaron.bsky.team shouting out how great using Osprey is for investigating! Go @roost.tools @julietshen.bsky.social #opensource #tssummit #trustandsafety

Anyone that has done content moderation before knows that emotional support/wellbeing tools are always developed secondary to the work (usually in response to burn out etc). Putting it first, ensures that moderators can take an approach to the work that actually centers their wellbeing.

Well. 😅

The lawsuit concerned allegations that Meta covered up its platforms’ impact on children's mental health and its knowledge of child exploitation online.

Meta is updating its child safety standards as countries across the globe consider social media bans for teenagers.

new from @katiemcque.bsky.social: @riana.bsky.social has talked about this at length, but the actionability of reports submitted by tech companies to NCMEC is a huge unsolved and growing problem www.theguardian.com/technology/2...

When Congress incentivizes over-reporting but leaves what to report up to platforms’ discretion (since for constitutional reasons the govt can’t tell private actors what to do), this is what you get. If NCMEC or LE can’t directly tell platforms what to (not) do, they drag them in the press instead.

West Virginia’s Anti-Apple CSAM Lawsuit Would Help Child Predators Walk Free West Virginia Attorney General JB McCuskey wants you to think he's protecting children. His press release says so. His legal complaint opens with the genuinely horrific line that Apple has, in internal communications,…

Through the EU's DSA and US litigation, new evidence is emerging about how social media platforms understand and address risks to minors. Peter Chapman and Matt Steinberg analyze what these parallel processes uncover about how platforms assess risk and design.

🫠

The American president is more insulated from accountability than a British royal. Our political system provides the elite with immunity. It has to change.

Amid the deals and demos at the India AI Impact Summit this week, the opportunity to shift the global debate about AI and what kind of world we are building appears to be lost, writes Tech Policy Press contributing editor Amber Sinha. www.techpolicy.press/indias-ai-su...

OK "god bless america" and then naming every country in the americas from south to north is absolute king shit

A must read fiction book honoring Fred in the best way. SO good. One of my fave books this year! Even more chilling with what is happening now.

New #research from Resolver “The Com” my takeaway: this isn’t one group it’s a growing network of online harms that spreads across platforms and mixes CSAM, self-harm, extremism & cybercrime. Addressing it requires coordinated action. No more silos! Read the full briefing here: ter.li/7ffbot

Incredibly rough to read but there needs to be more clarity in what platforms need to report to NCMEC & LE these annotations in the form are confusing.

It is very cool!!

This is huge news. I have spent the past 6 months wondering wtf was up with Amazon: they filed 380,000 AI-related CyberTipline reports to NCMEC in the first half of 2025. Turns out ALL of it was known CSAM they found by screening their AI training data. It's NOT AI-generated or AI-morphed CSAM.

production-grade tools for online safety CAN be built in the public, and CAN be collaboratively developed by engineers across different organizations. We're proud to announce Osprey and so grateful to all the contributors who made this possible! Release notes here: github.com/roostorg/osp...

I'm over a week late in reading this, but huge kudos to Charlie Warzel and Matteo Wong for doing the thing I thought nobody seemed to be doing, which is asking xAI's investors why they're giving money to the deepfake NCII/CSAM machine (spoiler alert: no answers): www.theatlantic.com/technology/2...

The EU is the latest regulator to investigate X’s Grok over AI-generated sexual abuse. Ramsha Jahangir reports on where governments stand — and what the fragmented response shows about the challenges of protecting children online across borders.

Our new memorandum of understanding with @ofcom.bsky.social builds on strong existing relationship and lays the groundwork for how we can assist Ofcom in its duties under the Online Safety Act. Read more on our website. www.iwf.org.uk/news-media/n...

@cassidyjames.com and I will be at #FOSDEM and will be demoing* Osprey in our talk Sunday morning: fosdem.org/2026/schedul... *Let's be honest, should we really test the wifi and live demo overlords first thing Sunday morning? Probably not. Probably pre-recorded.