Tech Policy Press

How to Manage Misinformation in Large Language Models A group of trust and safety professionals say models are ultimately only useful when they can be trusted.

The value proposition for AI systems ultimately hinges on trust, write Leah Ferentinos, Omri Tubiana, Arushi Saxena, J.J. Martinez-Layuno and Chris Miles. The necessary tools and frameworks are emerging to deliver it, they say.

Taiwan’s AI Basic Act Can Be a Model for Asia The new law lays down Taiwan’s “soft law” foundation for a development journey toward economic security through AI, write Wesley Chu and Charles Mok.

In December, Taiwan’s Legislative Yuan passed the AI Basic Act, and on January 14, it was signed by President William Lai Ching-te. The new law seeks to lay down Taiwan’s “soft law” foundation for a development journey toward economic security through AI, write Wesley Chu and Charles Mok.

EU Parliament Committee Confronts US Visa Bans Over DSA, Urges Commission Action The Committee on the Internal Market and Consumer Protection heard from four individuals facing US visa sanctions on Wednesday, reports Ramsha Jahangir.

Two months after Washington imposed visa bans on four individuals over the EU’s Digital Services Act, the European Parliament summoned them on Wednesday to account for what many MEPs see as a direct challenge to Europe’s regulatory authority, reports Ramsha Jahangir.

‘Community Benefits’ from a Hyperscale Data Center are a Mighty Tall Order Greg LeRoy directs Good Jobs First, a national nonprofit watchdog group on economic development issues.

Given the demands of hyperscale data centers, can these huge, extractive facilities ever be modified in ways that are actually win-win for communities where they are built? Good Jobs First director Greg LeRoy is concerned that community benefit agreements could get abused by data center developers.

In Age of Disruption, a Defense of Incrementalism Evan Selinger and Albert Fox Cahn are the authors of Move Slow and Upgrade: The Power of Incremental Innovation.

In their book, Move Slow and Upgrade: The Power of Incremental Innovation, Evan Selinger and Albert Fox Cahn argue society is over-fixated on disruptive innovation and needs to adopt an “upgrader’s mindset,” which should be applied whenever “disruptive changes would pose the greatest social risk.”

Privacy Regulators in 61 Countries Back Enforcement Against AI Deepfakes Regulators remind platforms that creation of non-consensual intimate imagery can constitute a criminal offence in many jurisdictions, reports Ramsha Jahangir.

61 privacy regulators across four continents say AI-generated nonconsensual intimate imagery violates data protection law. As investigations into Grok expand in multiple jurisdictions, the declaration places those probes within a coordinated global frame, reports Ramsha Jahangir.

No Digital Public Infrastructure Without Redress Trust at scale requires redress at scale, writes David Porteous.

Deployed at scale, all Digital Public Infrastructure (DPI) systems produce grievances. But will governments build systems capable of resolving them? Mistakes can exclude people from healthcare, banking, or welfare. Trust at scale requires redress at scale, argues David Porteous.

The EU’s Real AI Leverage Is Making Compliance the Path of Least Resistance How standardized evaluation, documentation, and incident reporting under the AI Act can become the global compliance default.

This spring, a frontier AI lab will prepare its first Safety and Security Model Report under the EU’s GPAI Code of Practice. Joel Christoph argues the real choice is whether to build for Europe alone or create compliance that other jurisdictions can use.

Done Data Work? You Might Have Worked With the US Military Without Knowing. As data workers organize around the world, they need to push for change on what appears to be a major ethical oversight, writes Krystal Kauffman.

The AI boom runs on invisible labor, writes Krystal Kauffman, a research fellow with the Distributed AI Research Institute (DAIR). Underpaid workers in Kenya, the US, and beyond are unknowingly training intelligence systems. This isn't just an ethics problem—it's a consent problem, she says.

Why Congress Should Step Into the Anthropic-Pentagon Dispute Stripped of its rhetoric, the Anthropic-Pentagon dispute resembles something relatively straightforward: a procurement disagreement, Daniel Castro writes.

The boundaries of military AI use should not be settled through ad hoc negotiations between a Cabinet secretary and a CEO, and the showdown between the Pentagon and Anthropic underscores that, ITIF's Daniel Castro writes

Surprise! The One Being Ripped Off by Your AI Agent Is You Laura MacCleery says policymakers must develop rules that matter before the vulnerabilities and harms of AI agents get much worse.

Moltbook, the Reddit-style forum for AI agents, is perhaps the most bizarre public AI experiment of late, writes Laura MacCleery. But the bigger story is that the clock is ticking for lawmakers to act before real harm from agentic systems scales up, says MacCleeery.

AI Impact Summit Commitments Must Counter ‘AI Unreality’ in Politics Democracies must invest in systems and institutions that help citizens discern fact from fiction, write GMF's Lindsay Gorman and Sharinee Jagtiani.

Policymakers and executives put up big commitments at the India AI Impact Summit, pledging to spend billions. But if democracy is to survive, leaders must also invest in the systems and institutions that help citizens discern fact from fiction, write the GMF's Lindsay Gorman and Sharinee Jagtiani.

How Brazil's AI Governance Vision Got Sidelined at the India Summit Brazil's multilateralism agenda was forsaken where it mattered most at India's AI summit, Tech Policy Press fellow Tatiana Dias writes.

Brazil arrived at the India AI Impact Summit with a clear mission to reinforce the need for digital sovereignty and multilateralism, but ran into a harsh reality, Tech Policy Press fellow Tatiana Dias writes

What Trump Did and Didn’t Say About AI in the State of the Union Zachey Kliger looks at President Donald Trump's new pledge around data center electricity costs and considers public opinion on issues he did not address.

President Donald Trump’s record-breaking 107-minute State of the Union address referenced artificial intelligence only three times, two of them in passing. Zachey Kliger looks at what he said and what he didn’t say about AI issues, including those related to jobs, children, and overall safety.

What the Father of Black History Can Teach Us About Technology Danielle A. Davis Canty, Esq., says that while technology may be miseducated, we do not have to be.

Historian Carter G. Woodson showed that when systems are designed and governed without the people most affected, inequality is not a mistake – it is the expected result. Technological inequality, in this context, is not surprising. It is predictable, writes Danielle A. Davis Canty.

How to Think About the Anthropic-Pentagon Dispute A conversation with Kat Duffy from the Council on Foreign Relations and Amos Toh from the Brennan Center for Justice.

The Pentagon wants AI that can fight wars without limits. Anthropic says there are lines it won't cross. This week, that standoff turned into an all-out confrontation. To unpack the news, Justin Hendrix spoke to Kat Duffy from the Council on Foreign Relations and Amos Toh from the Brennan Center:

What US Lawsuits Reveal About Platform Design That DSA Reports Don’t EU risk assessments and platform litigation in the US represent distinct approaches to governing and mitigating risks posed by social media platforms.

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.

‘Community Benefits’ from a Hyperscale Data Center are a Mighty Tall Order Greg LeRoy directs Good Jobs First, a national nonprofit watchdog group on economic development issues.

Given the demands of hyperscale data centers, can these huge, extractive facilities ever be modified in ways that are actually win-win for communities where they are built? Good Jobs First director Greg LeRoy is concerned that community benefit agreements could get abused by data center developers.

What the Father of Black History Can Teach Us About Technology Danielle A. Davis Canty, Esq., says that while technology may be miseducated, we do not have to be.

Historian Carter G. Woodson showed that when systems are designed and governed without the people most affected, inequality is not a mistake – it is the expected result. Technological inequality, in this context, is not surprising. It is predictable, writes Danielle A. Davis Canty.

How to Think About the Anthropic-Pentagon Dispute A conversation with Kat Duffy from the Council on Foreign Relations and Amos Toh from the Brennan Center for Justice.

The Pentagon wants AI that can fight wars without limits. Anthropic says there are lines it won't cross. This week, that standoff turned into an all-out confrontation. To unpack the news, Justin Hendrix spoke to Kat Duffy from the Council on Foreign Relations and Amos Toh from the Brennan Center:

What Trump Did and Didn’t Say About AI in the State of the Union Zachey Kliger looks at President Donald Trump's new pledge around data center electricity costs and considers public opinion on issues he did not address.

United States President Donald Trump’s record-breaking 107-minute State of the Union address referenced artificial intelligence only three times, two of them in passing. Zachey Kliger looks at what he said and didn’t say about AI issues, including those related to jobs, children, and overall safety.

How to Think About the Anthropic-Pentagon Dispute A conversation with Kat Duffy from the Council on Foreign Relations and Amos Toh from the Brennan Center for Justice.

The Pentagon wants AI that can fight wars without limits. Anthropic says there are lines it won't cross. This week, that standoff turned into an all-out confrontation. To unpack the news, Justin Hendrix spoke to Kat Duffy from the Council on Foreign Relations and Amos Toh from the Brennan Center:

How Brazil's AI Governance Vision Got Sidelined at the India Summit Brazil's multilateralism agenda was forsaken where it mattered most at India's AI summit, Tech Policy Press fellow Tatiana Dias writes.

Brazil arrived at the India AI Impact Summit with a clear mission to reinforce the need for digital sovereignty and multilateralism, but ran into a harsh reality, Tech Policy Press fellow Tatiana Dias writes.

Surprise! The One Being Ripped Off by Your AI Agent Is You Laura MacCleery says policymakers must develop rules that matter before the vulnerabilities and harms of AI agents get much worse.

Moltbook, the Reddit-style forum for AI agents, is perhaps the most bizarre public AI experiment of late, writes Laura MacCleery. But while the bots are posting and doing things more or less inanely, the clock is ticking for lawmakers to act before harms scale up, says MacCleeery.

AI Impact Summit Commitments Must Counter ‘AI Unreality’ in Politics Democracies must invest in systems and institutions that help citizens discern fact from fiction, write GMF's Lindsay Gorman and Sharinee Jagtiani.

Policymakers and executives put up billions in commitments at the India AI Impact Summit. But if democracy is to survive, leaders must also invest in the systems and institutions that help citizens discern fact from fiction, write the GMF Technology Program's Lindsay Gorman and Sharinee Jagtiani.

‘Community Benefits’ from a Hyperscale Data Center are a Mighty Tall Order Greg LeRoy directs Good Jobs First, a national nonprofit watchdog group on economic development issues.

Given the demands of hyperscale data centers, can these huge, extractive facilities ever be modified in ways that are actually win-win for communities where they are built? Good Jobs First director Greg LeRoy is concerned that community benefit agreements could get abused by data center developers.

What the Father of Black History Can Teach Us About Technology Danielle A. Davis Canty, Esq., says that while technology may be miseducated, we do not have to be.

Historian Carter G. Woodson showed that when systems are designed and governed without the people most affected, inequality is not a mistake – it is the expected result. Technological inequality, in this context, is not surprising. It is predictable, writes Danielle A. Davis Canty.

A Timeline of the Anthropic-Pentagon Dispute The dispute raises a variety of policy, legal, and ethical questions, and its outcome could set an important precedent.

Our timeline of the Anthropic-Pentagon dispute is updated to include the latest entry, President Donald Trump's statement on Truth Social "directing EVERY Federal Agency in the United States Government to IMMEDIATELY CEASE all use of Anthropic’s technology."

What Trump Did and Didn’t Say About AI in the State of the Union Zachey Kliger looks at President Donald Trump's new pledge around data center electricity costs and considers public opinion on issues he did not address.

United States President Donald Trump’s record-breaking 107-minute State of the Union address referenced AI only three times, two of them in passing. Zachey Kliger looks at what he said and didn’t say about AI issues, including those related to jobs, children, and overall safety.

Privacy Regulators in 61 Countries Back Enforcement Against AI Deepfakes Regulators remind platforms that creation of non-consensual intimate imagery can constitute a criminal offence in many jurisdictions, reports Ramsha Jahangir.

61 privacy regulators across four continents say AI-generated nonconsensual intimate imagery violates data protection law. As investigations into Grok expand in multiple jurisdictions, the declaration places those probes within a coordinated global frame, reports Ramsha Jahangir.