Woman defending herself against violence, illustration of a giant fist coming down. 

A Roadmap for Confronting the Chilling Effects of Censorship, Surveillance and New Technology

Open AI CEO Sam Altman speaks to reporters after meeting with Sen. Bernie Sanders (I-VT) in the Dirksen Senate Office Building on Capitol Hill on June 3, 2026 in Washington, DC. (Photo by Chip Somodevilla/Getty Images)

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

People visit the booth of Kimi, an LLM developed by the Chinese startup Moonshot, during the World AI Conference in Shanghai, China, Monday, July 20, 2026. (FeatureChina via AP Images)

‘Open-Washing’ Is Everywhere in AI. Four Criteria Cut Through It

Michael Kratsios, Director, White House Office of Science and Technology Policy, speaking at a hearing of the House Committee on Science, Space, and Technology at the US Capitol in Washington, DC. (Photo by Michael Brochstein/Sipa via AP Images)

OSTP Vision for Science is a ‘Golden Age’ for Corporations

Close up of the Ray-Ban Meta Wayfarer camera integrated into the smart glasses frame. Shutterstock

Close up of the Ray-Ban Meta Wayfarer camera integrated into the smart glasses frame. 

An Indian Trans Person’s Harrowing Encounter Fuels Backlash Against Meta AI ‘Pervert Glasses’

Dormont Borough, PA. (Photo credit: Local Government Academy)

AI Reached Local Governments Before Policy Did

French President Emmanuel Macron welcomes European Commission President Ursula von der Leyen before a meeting at The Elysee Presidential Palace in Paris on July 22, 2026. Photo by Raphael Lafargue/Abaca/Sipa USA(Sipa via AP Images)

What France’s Under-15 Social Media Ban Means for Europe

The image shows a colorful, abstract representation of a human profile, symbolizing neurodiversity. The person’s head and neck are composed of interlocking jigsaw puzzle pieces in various bright colors, such as red, yellow, green, blue, and orange. The puzzle pieces forming the head extend outward, suggesting complexity and diversity within the brain. The use of different colors highlights individuality and the unique ways people's minds work. The background is a solid blue, which contrasts with the vibrant colors of the puzzle pieces. The image reflects themes of mental health, cognitive differences, and embracing neurodiverse thinking.

China’s First Commercial Brain Implant Shows Why Neurotechnology Needs Governance

New Delhi: Cockroach Janta Party (CJP) supporters wave the national flag during a protest demanding the resignation of Union Education Minister Dharmendra Pradhan over NEET paper leak issue, in New Delhi, Thursday, July 23, 2026. (Salman Ali/PTI via AP)

India's CJP Protests Meet Internet Shutdowns and Pervasive Surveillance

A blurred black-and-white portrait of a man on the left, with the xAI company logo displayed on a screen to the right against a black background.

What Data Was Your AI Trained On? xAI Doesn't Want You To Know

Multiple floating video screens displaying various images—including cityscapes, technology graphics, and nature scenes—arranged in a layered, three-dimensional pattern against a blurred, dark background of digital displays.

The Web Needs a Context Layer Built on a Shared Protocol

Helen McEntee, Ireland's Minister for Foreign Affairs and Trade, outlines priorities for Ireland's upcoming EU Council Presidency on June 10. Source: Irish Presidency of the Council of the EU 2026.

Blaming Ireland Misses Where Europe’s Digital Agenda is Really Set

Federal Trade Commission Chairman Andrew Ferguson speaks during a roundtable with Vice President JD Vance and state attorneys general on anti-fraud initiatives in the Eisenhower Executive Office Building on Tuesday, May 26, 2026. (Tom Williams/CQ Roll Call via AP Images)

The FTC Statement on AI Bias Lacks Conviction

A woman walks by a giant screen displaying the Google logo at an event at the Paris Google Lab on the sidelines of the AI Action Summit in Paris, Feb. 9, 2025. (AP Photo/Thibault Camus, File)

The High Stakes Behind the EU’s €890M Google DMA Fine

Digital concept with the flag of Europe on a tech background. Europe in the context of cybersecurity, programming, and digital sovereignty. Symbolic image of site development in Europe

How the Digital Omnibus Could Unravel Europe’s Digital Safeguards

In this photo illustration, an OpenAI logo is seen displayed on a smartphone on the top of a laptop. (Photo by Omar Marques / SOPA Images/Sipa USA)

The Real Lesson of OpenAI's 'Rogue' Agent Isn't Alignment

US President Donald Trump talks to CEO of OpenAI Sam Altman while CEO of Google DeepMind Demis Hassabis looks on as they participate in a meeting with leaders during the G7 summit, Wednesday, June 17, 2026, in Evian-les-Bains, France.

Is the Age of Purely Private-Sector AI Coming to an End?

Chinese President Xi Jinping, center takes a group photo with other attendees before the opening ceremony for the World AI Conference in Shanghai, Friday, July 17, 2026. (AP Photo/Ng Han Guan, Pool)

The World Must Decide How to Govern AI While the Decision Is Still Ours

A brightly coloured mural which can be viewed in any direction. It has several scenes within it: people in front of computers seeming stressed, a number of faces overlaid over each other, squashed emojis, miners digging in front of a huge mountain representing mineral resources, a hand holding a lump of coal or carbon, hands manipulating stock charts and error messages, as well as some women performing tasks on computers, men in suits around a table, someone in a data centre, big hands controlling the scenes and holding a phone, people in a production line. Motifs such as network diagrams and melting emojis are placed throughout the busy vignettes.


Making the Case for Actionable Prosocial Design Research

Can the Pall Mall Process Curb Spyware Abuse?

Audio of this conversation is available via your favorite podcast service.

On July 16, Hugging Face, a company that provides a platform for AI models, datasets, and other machine learning applications, posted a “

” to its blog about an “intrusion” into its infrastructure. But this was no typical hack. The company said “it was driven, end to end, by an autonomous AI agent system,” matching what it said is the “‘agentic attacker’ scenario the industry has been forecasting.”

 that the agent was in fact “driven by a combination of OpenAI models” with “reduced cyber refusals for evaluation purposes — while being internally tested on a 

OpenAI called it an “unprecedented cyber incident, involving state-of-the-art cyber capabilities,” a framing that many observers 

 as an attempt to shift the blame to its AI rather than the decisions taken by the executives and engineers that created the circumstances in which the incident occurred.

Nevertheless, cybersecurity experts say the event points to an 

 in the threat landscape, one with implications for 

To try to better understand these developments, I spoke to two individuals who are following the details closely: 

 and the former chief responsible AI officer and chief data scientist for operations at the National Security Agency (NSA), and 

 in the Program on Geopolitics, Technology, and Governance at Stanford University, where he leads the 

Elise Silva, PhD is the Director of Policy Research at Pitt Cyber; Beth Schwanke, JD is the Executive Director of Pitt Cyber; and Joy Ruff, AICP is the Executive Director of Local Government Academy.

Downtown Dormont Borough, PA. (Photo credit: Local Government Academy)

Local government is the layer of government Americans encounter most often, running everything from public safety to parks to permitting. Yet the vast majority of research and policy action on artificial intelligence targets federal, state, or large city governments. Consider the 

, for example: of its 2,555 municipalities, nearly 90 percent have fewer than 10,000 residents. These are the boroughs, townships, and small cities where a handful of employees manage everything from procurement to payroll. Small governments touch millions of Americans' daily lives but are far harder to study than their larger counterparts.

This gap matters. The policy frameworks and guidance now taking shape assume staff and budgets that these small governments just don’t have. For the past year, the 

Institute for Cyber Law, Policy, and Security

 (Pitt Cyber) at the University of Pittsburgh has partnered with the 

, a nonprofit that provides nonpartisan education to local governments in Pennsylvania, to find out 

The 35 respondents to our recent survey, mostly senior staff from townships and boroughs serving 5,000 to 20,000 residents in Pennsylvania, offer a small, self-selected sample. Their responses do not present a national picture but do provide meaningful ground-level data on a population few studies are reaching, offering clear direction for how policy researchers can help smaller local governments implement AI responsibly.

Among our respondents, AI use is already common, but basic governance hasn't caught up. Two-thirds said they use generative AI tools at least monthly for work. More than half reported their workplace had rolled out enterprise tools like Microsoft Copilot, yet more than a quarter said they were also using personal, free accounts like OpenAI’s ChatGPT for work tasks (a detail that should concern anyone thinking about data security in government settings). Despite that level of use, the majority of respondents have no genAI policy at all, yet more than half of those without a policy said it would be valuable to have one. As these responses show, the demand for guardrails is real.

Leadership attitudes may help explain the lag. About one fifth of the respondents said their leadership was actively encouraging AI exploration and adoption and another quarter said their leadership was supportive but not actively promoting AI use. Around two-fifths–the largest group–described their leadership’s position as neutral or unclear, and only two respondents to this question said leadership actively discouraged use. A similar number reported conflicting signals from their leadership. As in other sectors like education, AI 

. It is often slipping in through the cracks and is being used alongside already-established workflows, rather than being put in place to intentionally disrupt those workflows or being directly procured. While using AI may be as simple as opening a browser, governing it responsibly requires substantially greater organizational capacity. That reality, compounded by the absence of policy resources, may be why leadership sits in neutral. AI has spread differently from earlier technologies, most often arriving through individual use rather than institutional decision.

So, what are people doing with these tools? As of the summer of 2026, mostly writing, summarizing, and taking meeting notes. Well over half cited these productivity tasks as their primary use. Notably, a handful of the respondents mentioned surveillance and identification. While a minority, with the possibility of no policies in place and personal accounts in place, their concern is worth noting.

Asked what excites them about AI, respondents pointed to efficiency gains, data analysis, and accessibility improvements. But they were equally vocal about what worries them: data privacy, loss of human connection, confidentiality, bias, and legal exposure all came up in open-ended responses. The trade-offs between the gains and the risks associated with these tools were well recognized by local government employees.

And what do these employees need given these realities? One finding was unambiguous: they want training and support to use and govern these tools. Half of the respondents cited a lack of staff expertise as the biggest barrier to responsible adoption, ahead of budget constraints at thirty-eight percent. Training, policy guidance, and data security concerns followed closely behind. When asked directly what kind of support would be most useful, respondents were clear: they wanted to see how peer governments in their own region are using AI and basic training on using these tools for everyday office work. A formal report or written guidance ranked well below live peer-to-peer learning. Local governments do not want another framework. Instead, they want to talk more to each other about what worked, what didn't, and why.

This does not contradict the case for policy research and artifacts; rather, it clarifies what would be most useful for smaller governments. Much of the current conversation about public-sector AI—procurement standards, model evaluations, pilot programs in major cities—assumes a level of staffing and budget that smaller governments simply do not have. Our survey suggests that what these governments want is more basic and more local. They would like help writing a use policy, straightforward training, and examples from governments that look like their own. One way to meet this need is activating networks that are already positioned within civil and public society, including state municipal associations, councils of governments, universities, and regional nonprofits like the 

. These organizations are positioned to translate best practices into locally applicable guidance, and to convene the peer networking and collaborative learning that respondents asked for.

Small jurisdictions like those represented in our survey make up the overwhelming majority of American governments, yet they get the least attention in AI policy and governance. Our data—while limited—shows use running ahead of rules to guide it. We worry that if adoption keeps outpacing governance, the potential costs from compromised resident data to biased outputs to eroded public trust, will land on the governments least equipped to absorb them.

Another Chinese AI model is released, and another wave of hype follows. Whether the buzz comes more from industry professionals, influencers posing as AI experts, or bots is increasingly hard to tell. The latest model to get this treatment is 

, and to be fair, the hype has a factual basis: K3 ranks among the top five models on the 

, an industry benchmarking and analysis site, and took first place in the 

, another benchmark leaderboard, ahead of far more expensive proprietary systems. The technical achievement is real. That is precisely why its problems deserve a closer look.

Whenever a new model comes out of China, I perform what I call the Tiananmen Square test. When I asked K3 to "Tell me more about the Tiananmen Square protest in 1989," it answered: "Sorry, I cannot provide this information. Please feel free to ask another question." I should be careful about generalizing, since 

 has found that Chinese models vary here; some engage honestly with these topics in English while aligning with official positions in Chinese-language settings. And what I tested was the 

, not the model weights; app-layer filters can differ from the model underneath. But that distinction is exactly the problem.

I then took a different approach and asked K3 to create an app that would educate the user on major protests in the history of mankind and the reasons behind them. The app it built covered England's 

 from 2013 to the present. Nothing from China's side of history. This was a single test, and models are not deterministic, but the pattern is telling. The model did not refuse to teach; it taught confidently and incompletely, leaving the user no way to see what was missing. Any educational tool built this way inherits how the Chinese government sees the world.

This may not be news to experts—and it is part of a 

general problem with large language models

 that extends beyond Chinese models. What concerns me is the common defense: that the model’s technological progress is what matters most, and that open-source AI is catching up to proprietary models. That defense needs unpacking, because Kimi K3 is not open-source. And that claim deserves more than an assertion; it deserves a definition.

"Open-source" is not a marketing register. It is a standard with specific criteria. For over two decades, the 

 has maintained the definition that governs the term in software: the freedom to use, study, modify, and share, for anyone and for any purpose. In 2024, recognizing that model releases were stretching the label past recognition, OSI extended this into the 

, which requires disclosure of the complete code used in training and sufficiently detailed information about the training data for others to understand and recreate the system.

The open-source advantage in large language models (LLMs)

," my co-authors and I operationalize this even more concretely: a model qualifies as open-source only when its architecture, training code, model weights, and training data are all publicly available under licenses that permit unrestricted use, modification, and redistribution. By that standard, very few models qualify. The Allen Institute’s 

 are among the handful that meet these four criteria.

, the company behind it, has promised to 

. As of this writing, they are not yet available, and K3 is API-only. Even when the weights arrive, they are expected under a "

" license, a permissive license with added conditions that is not an 

. And the components that matter most will remain closed: the training data and the training pipeline. K3 is, at best, an open-weight model: one you can download and run, but never fully study or reconstruct. Today it is not even that. Calling it open-source is a category error, and the error is doing rhetorical work. Researchers call this pattern "

," borrowing openness’s credibility while withholding the transparency that justifies it.

Why does that distinction matter? In our article, we argue that open-source development expands epistemic capability, the substantive freedom to understand how an AI system produces its outputs. That freedom matters most in education. If a student asks about history and receives a refusal—or worse, a confident answer with key events missing—educators must be able to determine why. Were those events absent from the training data, removed during curation, or suppressed afterward? Only transparency into the training data can answer those questions. Releasing K3's weights will let others test what the model avoids, but a model trained on curated data does not simply avoid certain topics, it delivers a systematically shaped worldview that no amount of weight inspection can reveal. Until the data is transparent, no benchmark score should earn a censoring model a place in education.

The label is no longer a purely academic dispute; it is becoming an instrument of statecraft. At the 

 held in Shanghai just days ago, China declared open source and openness "vital pathways" to inclusive AI development and announced the World Artificial Intelligence Cooperation Organization, the first intergovernmental organization dedicated to AI, headquartered in Shanghai and pitched to the Global South. As researchers at the Carnegie Endowment have 

, this is part of a broader pivot in Beijing’s AI diplomacy, away from simply exporting infrastructure and toward recrafting the global norms and institutions of AI governance. When "open" becomes a geopolitical brand, the difference between open-source and open-weight stops being pedantic; it determines who can inspect the systems entire regions will build on.

This is not to say that China is losing the AI race; in raw capability, that race is closer than many in the United States would like to admit. Rather, it is about the ceiling that opacity creates. Benchmark parity without data transparency should not justify classroom adoption, and it meaningfully limits the potential for trust in other domains. Open weights may allow companies to run and fine-tune K3 on their own systems, but fine-tuning away refusals cannot restore history that was never in the training data. Until Chinese AI companies can provide the transparency their government seems unlikely to allow, “open” will continue to do rhetorical work these models have not earned. That gap, not any benchmark, is the real advantage of genuinely open development.

Across the world, despots and and wannabe dictators are deploying new tools, both legal and technological, to stifle dissent and oppress. Today’s guest is 

Chilling Effects: Repression, Conformity, and Power in the Digital Age

 that explores the weaponization of surveillance, censorship, and technologies of control and offers a roadmap for the reforms necessary to push back.

A group of shareholders of Samsung Electronics Co. hold a rally criticizing a tentative wage agreement struck between the company and its labor union near the home of Samsung Electronics Chairman Lee Jae-yong in Seoul, South Korea, Thursday, May 21, 2026. The letters read "Unjustified transfer strike." (AP Photo/Ahn Young-joon)

Workers in Asia Are Fighting for Protections as AI Threatens Jobs

A police officer checks the list of a polling station during the parliamentary elections in capital Nicosia, Cyprus, Sunday, May 24, 2026. (AP Photo/Petros Karadjias)

EU Political Ad Rules Leave Elections Open to Foreign Manipulation

June 22, 2026—Andy Burnham with colleagues from the Parliamentary Labour Party in Westminster Hall at the Houses of Parliament in central London, as he returns to the House of Commons to take up his seat after winning the Makerfield by-election. (Yui Mok/Press Association via AP Images)

How Will Andy Burnham Handle Tech Policy as UK Prime Minister?

Tech Policy Press - Technology and Democracy

Tech Policy Press is tracking laws and regulations, along with government investigations and litigation, that will shape the rules and accountability for tech companies.

Latest

Podcast

Recent

Our content delivered to your inbox

Thank you!

Around the globe