Pre-release security review of frontier models is reasonable in principle, argues Michelle De Mooy. But a framework kept secret, with no published criteria or timelines, risks becoming a tool for executive discretion rather than governance. The Trump administration should provide answers, she says.

Five Questions the US Government Should Answer About Its Secretive Frontier AI Framework

Michelle De Mooy proposes a set of questions and standards against which whatever emerges from the Trump administration's should be judged.

techpolicy.press

Almost every LLM that reaches a broad audience is quantized—compressed to run cheaply. USC's Emilio Ferrara argues this compression is treated as a safety nonevent, when it should be treated as a change to the deployed system. This has policy implications, he says.

The Model You Audit Is Not the Model You Ship

Almost every LLM that reaches a broad audience is quantized—with under-appreciated ramifications for AI safety, writes Emilio Ferrara.

techpolicy.press

"The survey identified perceived downward pressure and a stunning 30 point decline when asked to compare their freedom to express themselves without fear of government interference or punishment ten years ago to their expectations for ten years from now."

John Sands@iohn.bsky.social · 1h ago

A new survey from @knightfoundation.org and @ssrsresearch.bsky.social reveals a disquieting decline in how free Americans feel to use the freedoms protected by the #FirstAmendment without government interference or punishment. 🧵

As opposition to data centers grows across the United States, the tech industry is "pulling from tried-and-tested playbooks" to bypass it, write Dean Jackson and Samantha Totoni—from nonprofit front groups to super PACs to tightly managed public forums.

Big Tech’s PR Playbook for Overcoming Data Center Opponents

The tech industry is pulling from tried-and-tested playbooks to bypass and securitize public opposition, write Dean Jackson and Sam Totoni.

techpolicy.press

CDT Europe's Christian Cirhigiri and Marie Seck welcome the EU panel's focus on children's rights and platform responsibility—but argue "safety by design" is left undefined, with no clarity on who decides when a platform is actually safe.

Does the EU's Special Panel Succeed in Charting a Third Path for Child Safety?

CDT Europe's Christian Cirhigiri and Marie Seck consider the Special Panel on Child Safety Online's July report.

techpolicy.press

"TikTok’s algorithm pushed him thousands of videos about suicide, sadness, hopelessness and loneliness, right up until he killed himself."

TikTok Withheld a Safety Feature From Millions. One Died by Suicide

A confidential document says an algorithm adjustment was kept from many users “by design,” as part of an experiment to measure engagement.

bloomberg.com

"The White House told top tech companies Tuesday that it would exempt certain AI systems from its plans for government vetting of new AI models, giving free tools known as “open weight” models a pass and focusing scrutiny on the latest technology from leading US companies like OpenAI and Anthropic."

White House will exempt ‘open’ AI systems from security review

Officials briefed tech companies Tuesday on a framework that would let officials vet cutting-edge systems for cybersecurity risks.

washingtonpost.com

What happens when users try to say goodbye to an AI companion? 🎧 CDT’s @mluria.bsky.social joins the latest @techpolicypress.bsky.social podcast to discuss how AI companions are designed to exploit our social instincts. Listen here:

When Users Say 'Goodbye' to AI

Laura Karpas on recent research from the Center for Democracy and Technology's Michal Luria and Harvard Business School's Julian De Freitas.

techpolicy.press

TLDR: Google be like “things that are actually harming people right now aren’t harms. The only harms that exist are the ones that might occasionally happen and look really bad so we need the protection of a regulator who we can blame for sleeping at the wheel when those things happen, too.”

Justin Hendrix@justinhendrix.bsky.social · 2d ago

Google's policy plan offers lawmakers a map of AI harm, writes Gregory Gondwe, an assistant professor at the University of Alabama and a faculty associate at Harvard’s Berkman Klein Center for Internet & Society. It puts frontier threats at the center,and more familiar structural harms at the edges.

The EU AI Act's transparency rules are live—but WITNESS's Bruna Santos and Jacobo Castellanos warn the exemptions could quietly hand the loudest producers of synthetic media a pass.

EU AI Act Transparency Rules Are Now In Effect. Was It A Missed Opportunity?

The EU must confront the technical and legal realities of the digital lifecycle, write WITNESS's Bruna Santos and Jacobo Castellanos.

techpolicy.press

Welp.

Hell Gate *subscribe today!*@hellgatenyc.com · 2d ago

This past weekend, former Mayor Eric Adams was in Wisconsin Rapids—population 18,654—trying to convince residents that they should allow a real estate developer to build a data center in the heart of their town. The public hearing did not go well. hellgatenyc.com/eric-adams-i...

Google's policy plan offers lawmakers a map of AI harm, writes Gregory Gondwe, an assistant professor at the University of Alabama and a faculty associate at Harvard’s Berkman Klein Center for Internet & Society. It puts frontier threats at the center,and more familiar structural harms at the edges.

Google's AI Governance Plan Draws the Boundaries of What Counts as Harm

Google seeks to shape regulation by deciding which harms deserve special attention and which appear familiar, writes Gregory Gondwe.

techpolicy.press

Tulane's Muira McCammon documented how Google turned off reviews for 90+ ICE detention facilities, erasing accounts of conditions inside. Google says disabling ICE facility reviews was "unrelated to a government request." But McCammon argues the effect is the same: dissent rendered invisible.

As Google Scrubs ICE Detention Reviews, Platform Power Serves State Interests

Google’s content moderation choices threaten to render accounts of treatment inside federal detention centers invisible, writes Muira McCammon.

techpolicy.press

Over a hundred academics asked Ireland to recuse itself from EU digital and tax files during its Council presidency. Article 19's Mark Dempsey and Columbia World Projects' Anna Marchese argue Ireland should welcome the scrutiny.

The Irish Presidency of the Council of the European Union Should Welcome Scrutiny from Academia

The question is whether Ireland is willing to take actions that inspire confidence in the impartiality of its decisions, write Mark Dempsey and Anna Marchese.

techpolicy.press

The European Commission gains full enforcement powers over AI companies on Aug. 2, including the ability to investigate providers, order corrective measures and impose fines. Joana Soares reports on what the first months of enforcement may look like.

Brussels Gains New AI Act Enforcement Powers as Autonomous AI Tests Regulators

The EU AI Act enters a new phase as Brussels gains enforcement powers over AI companies and introduces transparency rules for AI-generated content.

buff.ly

Y'all, when @tmz-news.bsky.social is giving America moral clarity about MAGA's Anthony Fauci obsession, things have clearly gone off the rails. My latest: www.altrightdelete.news/p/dysfunctio...

Dysfunction Junction

Fauci takes the 5th on the Hill. Here's what you need to know.

altrightdelete.news

👀

Dominic@domoffinland.bsky.social · 4d ago

this guy gets it

Some UK tech policy podcasting for your weekend – @justinhendrix.bsky.social talks through Andy Burnham's first moves on tech with @gavinfreeguard.com, @rachelcoldicutt.bsky.social and me. Available here, and wherever you get your pods.

Taking Stock of the UK Tech Policy 'Muddle'

A conversation with Tech Policy Press fellow James Ball, State of the Future's Gavin Freeguard, and Careful Industries' Rachel Coldicutt.

techpolicy.press

Digital ID scrapped. DSIT shuttered. AI policy reshuffled. New UK Prime Minister Andy Burnham is already making moves to change the government’s approach to tech policy. I asked James Ball, Gavin Freeguard, and Rachel Coldicutt to help me make sense of the "muddle."

Taking Stock of the UK Tech Policy 'Muddle'

A conversation with Tech Policy Press fellow James Ball, State of the Future's Gavin Freeguard, and Careful Industries' Rachel Coldicutt.

techpolicy.press

One of the most amazing light phenomena I've ever seen. Just now on Vancouver Island.

The Trump administration tried to use visa denials and deportation to silence researchers who study social media. A court just said no. Sunday Tagliasacchi and Elisabeth Snyder, former legal externs at the Knight First Amendment Institute at Columbia University, break down the CITR v. Rubio ruling.

Trump Cannot Use Immigration Law to Silence Technology Researchers

Sunday Tagliasacchi and Elisabeth Snyder detail an initial court victory for the Coalition for Independent Technology Research in CITR v. Rubio.

buff.ly

"The attack has little precedent, experts said, but has long been the stuff of nightmares and sensationalized Hollywood thrillers: an apparent cyberattack by a foreign power during a time of war that could, at least in theory, jeopardize the health and safety of Americans."

Scope of Hacks on U.S. Water Supply Widens as Evidence Points to Iran

Michigan and Minnesota are among at least seven states coping with cyberattacks aimed at disrupting water systems nationwide.

nytimes.com

DHS is spending billions on domestic surveillance and data tools capable of targeting anyone, with little oversight or transparency. Congress must investigate how these powerful technologies are bought and used. bit.ly/4wx8EZO

DHS’s Domestic Surveillance, Data Collection, and Analytics Tools

Fueled by a budget boom, the Department of Homeland Security is expanding its technological arsenal, with little oversight or transparency around how these tools are used.

brennancenter.org

ICYMI: Hugging Face said a July breach was "driven, end to end, by an autonomous AI agent system." That system, it turned out, was operated by OpenAI. Vinh Nguyen and Graham Webster joined the Tech Policy Press podcast to discuss what the incident may mean for the geopolitics of AI governance.

How the OpenAI-Hugging Face Hack May Affect the Geopolitics of AI Governance

A conversation with Council on Foreign Relations senior fellow Vinh Nguyen and Stanford researcher and DigiChina Project lead Graham Webster.

techpolicy.press

Merici Vinton from the Center for American Progress & Lynn Overmann from Georgetown's Beeck Center extract five principles from NYC's new Public Interest Technology Crew that they argue should be standard practice for digital capacity initiatives everywhere.

Cities and States Can Replicate NYC’s Ambitious Public Interest Tech Crew

Merici Vinton and Lynn Overmann extract five principles they say should become the standard operating procedure for digital capacity initiatives everywhere.

buff.ly

Dynamite blasting and nuclear waste are treated as "ultrahazardous"—strict liability applies no matter the precautions. Reflecting on the OpenAI-Hugging Face incident, Keith Porcaro of Duke Law argues advanced AI belongs in the same category.

Advanced AI Is Ultrahazardous. Let’s Treat It That Way

Advanced AI creates some degree of risk to digital infrastructure that cannot be fully mitigated, writes Keith Porcaro.

techpolicy.press

13. Hence our dismal conclusion: rather than finding that LLMs free us to do a better job of what we were doing before they came along, they shift scientific incentives (and the playing field of academic competition) in ways that compel us to do more and more, faster and faster, less and less well.

A network of Telegram "local news" channels has no newsroom behind it. Democracy Reporting International’s Ognjen Karadzic & Daniela Alvarado Rincón trace how it grew across 19 languages undetected—and argue the EU should reconsider its light-touch approach to Telegram.

What a Growing Telegram Disinformation Network Reveals About the DSA

Ognjen Karadzic and Daniela Alvarado Rincón analyze their research on a Telegram disinformation network and what it means for the EU's Digital Services Act.

buff.ly

The fact that it’s poisoning Black people is not seen as a negative to these folks.

SpaceX won’t remove all of xAI’s unpermitted turbines for another year | TechCrunch

SpaceX is building a new power plant for xAI's Colossus data centers, but it won't remove existing, unpermitted turbines for many more months.

techcrunch.com

Mozilla CTO Raffi Krikorian says the real question for builders isn't which AI model they are using—it's whether they could leave for a different one. The door's still open to choose a different path—for now.

Is Open-Source AI Really the Dangerous Path?

The real question for builders right now isn’t which model you’re using. Rather, it's whether you could leave for a different one, writes Raffi Krikorian.

buff.ly