Google CEO Sundar Pichai talks before a dinner with President Donald Trump and Japan's Prime Minister Sanae Takaichi in the State Dining Room of the White House, Thursday, March 19, 2026, in Washington. (AP Photo/Julia Demaree Nikhinson)

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

A view of the Berlaymont European Commission building in Brussels, Belgium on July 20, 2026. (Photo by Jakub Porzycki/NurPhoto via AP)

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

June 6, 2026: Protesters and officers with the Department of Homeland Security clashed outside the Metropolitan Detention Center in Los Angeles on the first anniversary of increased US Immigration and Customs Enforcement (ICE) raids across Los Angeles County. (Photo by Ted Soqui/Sipa via AP Images)

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

Irish Minister for Foreign Affairs Helen McEntee delivers a speech at Dublin Castle in Dublin, during the opening ceremony of Ireland's EU presidency on Wednesday July 1, 2026. (Niall Carson IRISH Press Association via AP Images)

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

Illustration of a Tamagotchi with the OpenAI logo depicted on its screen. Tech Policy Press/Shutterstock

Illustration of a Tamagotchi with the OpenAI logo depicted on its screen. Tech Policy Press/

When Users Say 'Goodbye' to AI

A group of people walking around at the National Conference of State Legislatures. 

The 2026 National Conference of State Legislatures in Chicago. Photo by Justin Hendrix.

At NCSL, Geoffrey Hinton Warns State Lawmakers Not to Buy Big Tech's Anti-Regulation Pitch

UK Prime Minister Andy Burnham speaking during a visit to a family center in Sheffield, South Yorkshire, on Friday, July 31, 2026. (Annabel Lee-Ellis/PA Wire POLITICS)

Taking Stock of the UK Tech Policy 'Muddle'

Screenshot from a Google announcement depicting its launch of a Nano Banana 2 feature in Google Earth. 

Google Earth AI Fiasco Underscores Why Tech Firms Must Listen to Outside Experts

Mayor Zohran Kwame Mamdani and Chief Technology Officer and Office of Technology & Innovation (OTI) Commissioner Lisa Gelobter announced the launch of "Public Interest Technology (PIT) Crew," a new initiative to deploy teams of technologists alongside City agencies to tackle public problems with in-house digital solutions at Coney Island in New York City on July 13, 2026. Credit: Katie Godowski/MediaPunch /IPX

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

This image is a collage with a colourful Japanese vintage landscape showing a mountain, hills, flowers and other plants and a small stream. There are 3 large black data servers placed in the bottom half of the image, with a cloud of black smoke emitting from them, partly obscuring the scenery.

Deborah Lupton, "Servers in a Landscape" / 

Is Open-Source AI Really the Dangerous Path?

Telegram on App Store displayed on a phone screen and French flag displayed on a screen in the background are seen this illustration photo taken in Krakow, Poland on December 3, 2023. 

(Photo by Jakub Porzycki/NurPhoto via AP)

What a Growing Telegram Disinformation Network Reveals About the DSA

A woman waves a European Union flag in the wind.

The ‘Sovereignty’ Narratives That Don’t Help Europe

Kherson, Ukraine—Lenka Klicperova, journalist and photographer, covers the war in Ukraine. October 19, 2022 Jose Hernandez/Shutterstock

Kherson, Ukraine—Lenka Klicperova, journalist and photographer, covers the war in Ukraine. October 19, 2022 Jose Hernandez/

What It Now Takes to Establish a Fact

Editorial collage in black, white, and blue showing a megaphone emerging from a laptop screen and a hand holding a microphone, surrounded by ribbons reading 'NEWS' and scraps of newsprint.

How Cyberattacks Became a Censorship Tool

OpenAI founder and CEO Sam Altman speaks with reporters as he departs a meeting with Sen. Raphael Warnock (D-Ga.) on Capitol Hill on July 29, 2026. (Francis Chung/POLITICO via AP Images)

The OpenAI–Hugging Face Incident Demands Urgent Congressional Oversight

Nine small images with schematic representations of differently shaped neural networks, a human hand making a different gesture is placed behind each network.

Can AI Agents Be Aligned with Human Rights?

The X logo is being displayed on a mobile phone with the European Commission logo seen in the background, in this photo illustration. Taken in Brussels, Belgium, on 8 December 2025. (Jonathan Raa / Sipa USA) (Sipa via AP Images)

X's Data Access Remedies Are a Win for Researchers Only If They Last

A person holds a Mastercard debit card near a mobile phone with a contactless POS terminal in Athens, Greece, on February 15, 2026. (Photo by Nikos Pekiaridis/NurPhoto via AP)

(Photo by Nikos Pekiaridis/NurPhoto via AP)

Pakistan’s Banking System Holds a Lesson for Social Media Age Verification

A view of the Barlaymont European Commission building in Brussels, Belgium on July 20, 2026. (Photo by Jakub Porzycki/NurPhoto via AP)

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

The EU flag with an image of the X logo. 

What X's Corrected Action Under the DSA Fixes on Data Access, and What's Left Open

Muira McCammon is an assistant professor at Tulane University.

June 6, 2026: Protesters and officers with the Department of Homeland Security clashed outside the Metropolitan Detention Center in Los Angeles on the first anniversary of increased US Immigration and Customs Enforcement (ICE) raids across Los Angeles County. (Photo by Ted Soqui/Sipa via AP Images)


In June, United States Immigration and Customs Enforcement (ICE) agents tracked down 

 to the Syracuse Central Library, where she was working as a poll worker. They handed her a written warning and told her that her Instagram account had made threats against an ICE agent and was in violation of federal law. On January 8, Gonyea had posted the 

BREAKING: The ICE agent who shot and killed Renee Good in broad daylight has been identified as Jonathan Ross by the Minnesota Star Tribune. I think today is a good day for Jonathan to be indicted!

Under the Trump administration, ICE agents have been on a campaign to erase such critical statements, and lately, they’ve been getting help from Big Tech. Last year, I 

 that Google was turning off access to reviews and ratings of ICE detention facilities and many other federal components. By August 2025, the tech company had removed reviews for over 90 facilities, forbidding people from further commenting on them and eliminating access to previous comments. Last month, of that original group of detention facilities, only four still allowed reviewing and rating: the Farmville Detention Center, the Chase County Detention Center, the CNMI Department of Corrections, and the Grayson Center Detention Center. Previous comments, now gone from the internet, discussed conditions inside the camps and represented ongoing efforts to hold ICE accountable. I wrote about this in a 

 published last month in the journal New Media & Society.

, Google said that turning off access to reviews and ratings for ICE facilities was “unrelated to a government request,” and separately, in a statement to 

 a Department of Homeland Security spokesperson said that ICE had not requested any reviews of detention facilities be removed. After my work 

, Google turned off reviews for the four detention facilities, prompting concern from people like 

, whose husband Fernando remains detained at ICE’s Farmville Detention Center.

However, there’s a bigger issue at play here. Many of the communication channels where people are going to speak up against the US federal government are now managed by private technology companies. Google’s content moderation policies and how they may relate to 

 and reviews of individual US federal agencies are not being clearly disclosed to the public.

Google, for its part, has not announced any specific policy on the moderation of federal listings in Maps and only recently acknowledged—after 

 was published—that it was moderating comments related to ICE detention facilities. It is not yet known how aware federal agencies are of the knowledge panels that relate to their work and mission. How or if government workers are involved elsewhere in the US administrative state in the maintenance and editing of individual knowledge panels, which can be “

” by organizations, is unclear. If they are suggesting edits to these panels, it could make such efforts subject to the US Federal Records Act and other records management laws.

Google makes leaving reviews relatively easy, and at surface level, it might seem less risky to write on an ICE knowledge panel than to 

 But, the Trump administration has made it clear that it is coming after critics of the US carceral state, filing 

 who call out its inhumane behavior online and 

, it’s possible that anyone leaving a Google review critical of ICE, or other federal agencies housed outside the Department of Homeland Security, could face repercussions.

Google reviews: A medium for contesting institutional abuses and ICE

While online reviews may seem trivial, they are sometimes a key source of information. As journalists have struggled to see firsthand what’s happening within detention facilities, they’ve looked online for narratives of life inside the camps. In 2025, reporters at 

published a detailed forensic analysis, attempting to map the scope of the US immigration industrial complex. Their investigation drew on satellite images of facilities from Google Earth and excerpts of Google reviews. Other journalists have turned to Google reviews of ICE facilities, including Stephanie Koithan with 

, who documented conditions at the South Texas Processing Center in Pearsall.

Last year, in August 2025, I collected over a thousand Google reviews of 26 individual detention facilities that were still accessible. By then, Google had already begun turning off reviews for most of the facilities listed on the Department of Homeland Security’s official index of active ICE facilities. Today, only four of those facilities still allow for reviews to be posted. When my own scholarly article reporting my findings was published in June, the number of publicly available Google reviews had fallen from four digits to a mere 55. (Recently, 

interviewed a formerly detained person who gave the Adams County Correctional Center outside Natchez, Mississippi one star.)

This does not come as a total surprise. In 2023, Google, via one of its 

 hinted at the company’s practice of moderating reviews of prisons:

 where we have found user contributions to be consistently unhelpful, harmful, or off-topic. This includes places that people go to without choice or places only accessible to people stationed or assigned there — such as police stations and prisons.

Google has demonstrated a willingness to suppress the speech of people who seek to question the US administrative state. If the company is willing to quietly remove critical reviews of ICE detention facilities, what other types of expression might it be willing to erase? What about taxpayers who want to criticize the work of the 

? Or Americans and others around the world who might yearn to reminisce on the legacy and global impact of the now gutted 

United States Agency for International Development

? And what of those who might want to reflect on their interactions with the 

US Department of Justice’s Office of the Pardon Attorney

? At the moment, reviews for all of these entities are still up, but it is uncertain how the tech company might interpret its own commercial and political prerogatives in the coming years.

Google search results page for "Henderson Detention Center Reviews," showing an empty Reviews section with a MapQuest listing and a "People also ask" panel on the left, and a business info card for Henderson Detention Center in Clark County, Nevada, on the right with a photo, map, and contact details. The screenshot indicates "Posting is currently turned off."

What are reviewers of ICE detention facilities saying? Of the hundreds upon hundreds of reviews that I read, only a few bore obvious traces of internet ‘

 “#abolishice.” In many instances, self-identified family members of detained people wrote Google reviews to attest to the communication struggles they faced while attempting to contact their loved ones. As one person explained, “This place does not know what it’s doing [

] you called ask questions and you get transferred to phone numbers that don’t even exist.” Others explicitly expressed a desire to access contact information for detention facilities and their officials: “Can someone give me information about the email address of the officer in charge?” Some reviewers put their complaints more succinctly: “Learn to answer your damn phones!”

Google reviews panel for Adams County Correctional Center at 20 Hobo Fork Rd, Natchez, MS, showing a 2.4-star rating from 13 reviews and a featured 1-star review from user Amr Yousuf describing detention conditions as a nightmare and criticizing the facility as profit-driven.

For others, the challenge was not making calls but rather receiving them. As one self-identified former girlfriend of a detainee lamented, “I keep getting calls from this place and I’m not sure if my ex is trying to contact me.” One reviewer describing herself as the wife of a detained man similarly lamented, “If I could give 0 zeros I would I’ve called them over 1,000 times to know about my husband and no one answers.” Many of the comments criticized ICE communication practices fraught with uncertainty and disrespect.

What the comments reveal collectively is an institution that apparently doesn’t want to listen or dialogue with people, from the formerly detained to civically-engaged Americans.

What kind of democracy does Google want for the US?

Today, it would appear that Google’s commitments to free expression are not calibrated to the interests of people questing for accountability or federal oversight, but rather to the 

 doing the Trump administration’s bidding. If the company, through its content moderation practices, is helping to protect ICE’s institutional reputation through erasure and archival annihilation, what, if any, commitment does it have to people leaving reviews on other government institutions’ knowledge panels?

Historically having an informed citizenry has been seen by many as a prerequisite for a democracy, but the matter of where people can go online to hold the US federal government accountable has involved an increasingly complicated set of sociotechnical choices. Under the US Administrative Procedures Act, people can submit comments and provide feedback to new regulations proposed by individual US federal agencies (notably in 2018, over 

individuals left public comment for ICE, when it was proposing a rule change to how it treated unaccompanied children), but there is no requirement that government officials maintain—or listen—to other forums online.

These circumstances leave a number of questions open:

As formerly detained people reflect on their experiences, where will they be able to go in the future online to share their stories and try to hold the US federal government accountable, and who will control those channels?

Were a federal sub-agency to claim a Google Business Profile, the case could be made that a Google Knowledge Panel constitutes a public forum for purposes of the First Amendment and individuals whose reviews have been erased could have a potential case against the US federal government. How might an American court respond to a complaint from someone whose speech has been erased?

While a Knowledge Panel will generally fall outside the scope of the US Federal Records Act, editing the panel through Google's tools or responding to reviews might make its metadata qualify as a government record under the same logic that the National Archives and Records Act applies to agency use of 

 and other third-party platforms. Might legislators in Congress consider amending the US Federal Records Act to make it more inclusive of accounts of conditions at these fraught sites?

Tech companies have proven that if the right political moment comes around, they are willing to unearth the archive, as Twitter did in the restoration of President Donald Trump’s Twitter account. If a new president or political environment were to arise, could Google be pressured to turn back on reviewing and resurfacing past comments? Might Republicans and Democrats be united in this initiative, given the ideological diversity of narratives that have been posted on ICE facilities’ Google Knowledge Panels?

Could other countries’ governments regulate this practice and open up the potential for reviewers outside the US to contribute to these forums without the same threat of erasure?

Ultimately, Google reviews remain one of the main public-facing portals where people can complain, protest, and recount the intolerable conditions faced within many ICE detention facilities. Google’s content moderation choices threaten to render these narratives invisible.

On July 9, days after Ireland assumed the 6-month rotating presidency of the 

, over fifty academics (the number is now closer to a hundred) published an 

 asking the country to recuse itself from chairing meetings on digital and tax files given its “insurmountable conflicts of interest”. Unsurprisingly, the recusal request was not met. In fact, Ireland’s official response was a mixture of 

 and denial, aptly encapsulated in Taoiseach Micheál Martin’s insistence that his country would be an “

 and many others felt compelled to sign the letter must be taken seriously, not least because their concerns do not appear in isolation. Leading Irish economists like 

 have expressed similar unease about the country’s financial dependence on the foreign-owned multinationals it is responsible for regulating. 

Worse still, shortly after the letter appeared, photos emerged of Ireland’s Minister for Enterprise, Peter Burke, 

 at the Dublin residence of the US ambassador. Their aim? To discuss “the digital agenda for Ireland’s EU presidency,” according to the Irish Independent.

Not to worry though, we have been assured that Ireland can still act as an honest broker.

An ever-present tension: when economic reliance clashes with regulatory independence

Google, Meta, Apple, TikTok, OpenAI, X, and Microsoft have their EU headquarters in Ireland and the country’s regulators, most prominently the Data Protection Commission (DPC)—by virtue of the ‘country of origin’ principle—has become the primary watchdog for the tech sector. Unfortunately, 

, Ireland’s record of enforcement is not good. Defenders will point to the €4 billion in GDPR fines imposed by the DPC to assuage concerns. This, however, paints an incomplete picture. In the Meta case, which led to a 

€1.2 billion fine against the company in 2023

DPC at first recommended no financial penalty

. Only after the European Data Protection Board (EDPB) intervened did the Irish agency reverse course. This is not an isolated incident. It has regularly faced accusations of leniency, 

 of its GDPR decisions that have been subsequently overruled by the EDPB.

Days before the publication of the academics’ letter, which lays out a litany of troubling enforcement actions by the country, more facts have shed light on Ireland’s role as the hub of Big Tech’s EU business and the custodian of Europe's tax base.

On June 30, Microsoft published its first 

—an EU-mandated breakdown of its profit and income tax by country. Ireland accounted for 

 of the revenue and pre-tax profit reported in Microsoft's country-by-country filing. That’s $196 billion in total. To put that in perspective, the company reported thirty times more revenue in Ireland than Germany, fifty-one times more than France, one hundred fourteen times more than Italy and a hundred seventy two times more than Spain. With $47 billion in pre-tax profits in Ireland, the company generated roughly $7 million in profit per Irish employee, compared with an average of $375,000 per employee across its global workforce. These figures are difficult to justify, which might be one reason why Ireland voted against legislation mandating their public disclosure in 

. Considering Ireland’s own opposition to tax transparency, calls for the Irish Presidency to step aside from tax files are hard to dismiss.

It's easy to see why this might be a hard pill to swallow. Today, Ireland’s economic transformation is regarded as one of the EU’s greatest success stories, with a 

 that stands at more than double the bloc’s average. Yet, within the country, maintaining this momentum is framed as a fraught choice: protect the economic interests of American tech and pharma companies or jeopardize thousands of jobs and risk a return to the economic 

, when GDP per capita was less than 60% of the EU average and under one-quarter of that of the US. 

This approach is admittedly simplistic, as the long-term unsustainability of Ireland’s economy on corporate tax income is now 

. It also ignores the public costs of failing to hold tech companies to account. From 

 to knowingly deploying products that harm children, the damage that flows from weak oversight is no longer hypothetical. Tech entrepreneur and 

 founder Marie Potel-Saville recently wrote of how platforms use “

” to target kids, citing internal company research describing these methods as an “addict’s narrative”. It is long past time for Ireland to take its responsibility in mitigating these harms more seriously.

Ireland should treat this moment of renewed scrutiny as an opportunity to demonstrate real leadership and assert its regulatory power vis-à-vis Big Tech. As the EU prepares to rework its digital and tax rulebooks and as 

 continues to mount against European officials on both fronts, this Presidency can, if Ireland’s leaders so choose, also model how public institutions ought to respond when faced with conflict-of-interest accusations.

Instead of accepting the familiar argument that if Brussels’ EU institutions struggle to withstand Big Tech lobbying efforts, Ireland stands little chance of systematically rooting out illegitimate influence, Ireland should act today and build on the expectations of its European partners by reinforcing enforcement milestones, strengthening disclosure requirements, and embracing accountability mechanisms that identify and guard against both actual and 

conflicts of interest. Crucially, there are 

clear ethical guidelines that deal directly with the latter

. The mere appearance of a conflict is treated akin to an actual conflict, because, when it comes to public authority, these optics are just as damaging to institutional trust.

Ireland should therefore welcome new accountability guarantees. In this environment of rapidly moving legislation, and in a sector flooded with corporate money, no safety measure is superfluous.

There are ample opportunities to show resolve. A 

 in Competition Policy International shows that a booming re-identification industry is routinely undermining the intent of the GDPR’s core privacy protections. We invite the Irish DPC to investigate this industry as a matter of urgency.

The European Parliament can be a crucial ally

With its history of establishing committees of inquiry, the European Parliament would be a natural ally in this effort. These committees have reassured EU citizens that their elected representatives and public servants work in the public interest—not for private business or foreign powers—by exposing occasions when they have failed to do so. Examples include the 

 scandal which implicated then Commission President Jean-Claude Juncker, the 

 following the Panama Papers revelations and the 

 established to investigate the use of Pegasus spyware. 

The relationship between Big Tech firms and the institutions responsible for regulating them on behalf of the European Union warrants the same level of parliamentary scrutiny. As the EU prepares to debate the 

 process under the Irish Presidency, a dedicated DIGI committee could document and monitor every stage of the legislative deliberations, gather needed testimony from regulators, industry representatives, and independent experts, and issue recommendations for institutional reform or referrals to the appropriate authorities should it reveal evidence of maladministration or breaches of EU law.

, pressure on the EU’s digital rulebook extends beyond Ireland. Across Europe, policymakers are insisting on the rollback of regulatory safeguards dealing with today’s dominant digital platforms and AI companies for the sake of “competitiveness”. Nowhere is that pressure more acute than in Ireland.

The issue is not whether Ireland can completely insulate itself from corporate lobbying—no democracy can—but whether it is 

 to manage institutions and make public commitments that inspire confidence in the impartiality of its decisions. By embracing parliamentary oversight and accountability, this Irish Presidency has the occasion to prove that Europe's digital rulebook is enforced fairly, with vigor and beyond reproach.

Yesterday, the transparency rules in the European Union’s AI Act 

. The date represented more than a regulatory deadline; it marked a pivotal shift in the governance of synthetic content. We are now witnessing a collision between the ambitious promise of transparency and the complex, often dangerous reality of how content behaves in the wild. This moment forces a difficult question: are we building systems that genuinely protect the public, or are we merely masking systemic risks with superficial labels?

At the center of this transition lie Article 50’s transparency obligations and the newly minted 

, which aims to provide developers and deployers with a roadmap for identifying AI-generated media, chatbots, and biometric systems. Yet, the foundation of this effort—the CoP and its 

—reveals a fragility. There is a persistent gap between technical mandates and the reality of how content lives, breathes, and spreads online. To be truly effective, these regulations must move beyond the safety of binary labels and confront the technical and legal realities of the digital lifecycle, ensuring that transparency is as durable as the content it describes.

The first section of the Code sets rules for Providers of Generative AI Systems, and further develops the requirements present in Art. 50 (2) and (5). Overall, this section is responsible for outlining transparency and labeling obligations for providers of generative AI systems, focusing on four core commitments: implementing a multi-layered approach to marking AI outputs (e.g., digitally signed metadata and watermarking), providing free and functional detection tools that adhere to EU privacy laws, ensuring quality standards like robustness and interoperability, and establishing rigorous compliance, monitoring, and testing processes. Lastly, the final text also requires providers to make best efforts to preserve metadata markings and prohibits their removal, while offering proportionate compliance pathways for SMEs and startups to ensure that transparency measures remain effective and reliable across the content lifecycle.

While we welcome the progress made in the Code's provisions on metadata, watermarking, detection, interoperability, and privacy-preserving detection services, some important gaps remain. The main gap is that provenance information (Section 1, Measure 1.3) is designated optional and ‘encouraged’ while a binary AI generated marking (Measures 1.1.1 and 1.1.2) is all that the code requires of signatories which limits transparency across the content lifecycle.

On open-weight models, the code's reliance on voluntary cooperation, rather than structural marking obligations on model releases, leaves a significant gap: compliance with regulation cannot be enforced once weights are released and models are fine-tuned or stripped of provenance signals downstream.

Public reporting requirements and civil society oversight remain limited under the Code, a gap that weakens its overall trust and accountability mechanisms.

Finally, user-facing transparency benefits from stronger media literacy and clearer provenance information, ensuring that transparency measures are meaningful and understandable for the people they are intended to serve. However, the Code makes this optional and encouraged, stopping short of the mandatory obligations these measures would need to be effective.

Section 2 mandates transparency for deployers of AI systems, and it is the part of the Code closest to our own work. We welcome the disclosure obligations it sets, but from what we see through the 

 we would add one caution: a label is not the same as protection, and it should not be mistaken for one. For the deepfakes that cause the gravest harm, above all non-consensual intimate imagery, disclosure does little for the person targeted, and the recourse for genuinely illegal content lies elsewhere, in the 

 and, for image-based sexual abuse, in the 

Directive on combating violence against women

 (Directive (EU) 2024/1385). The Code works best read alongside those frameworks, and with safeguards so that recourse does not tip into the over-removal of satire, dissent or human rights documentation.

Where the Code could go further is in the strength of the label itself. The most informative layer, the one that would tell a viewer what was actually changed, is only encouraged rather than required (Section 2, Measure 1.1), and the marking behind the label carries no robust provenance, the same gap we note for providers in Section 1 (Measure 1.3). Because labels are also easily lost as content is edited, re-encoded and re-shared, more durable and provenance-rich marking would make disclosure far more meaningful in practice. The Code does, though, deserve credit for making the accessibility of disclosures binding, so that they work with screen readers, high contrast and audio or haptic cues.

Two things are worth watching as the rules take effect. The law-enforcement exemption, which lifts disclosure where a use is authorized to detect, prevent, investigate or prosecute a criminal offense (reflecting Article 50(4)), is the kind of state-use carve-out that can travel badly when copied into jurisdictions with weaker safeguards, at a moment when even official accounts, including the 

, are posting unlabeled synthetic imagery. And detection, which enforcement ultimately leans on, remains unreliable and uneven across languages and faces, as our 

 shows, so it should not be treated as a dependable backstop.

Do the guidelines water down the obligations?

The guidelines clarify some of the concepts in the Code, set out the applicable exemptions and define what falls out of scope, and explain how compliance with the Article 50 obligations may be demonstrated, including by signing the Code. They are a useful addition, but they also open friction points with Article 50 and the Code.

The first is vagueness in the key exemptions. The "obviousness" exception for interactive AI, and the line between exempted "standard editing" and regulated "substantial semantic alteration", both rest on subjective, case-by-case judgements, which creates real uncertainty for developers, deployers and platforms that blend ordinary editing with generative tools.

The second is an internal tension on watermarking. The guidelines acknowledge that universal, provider-agnostic watermarking and detection do not yet exist, especially for text, while the Code requires synthetic outputs to be machine-readable and detectable through "effective, interoperable and reliable" solutions. As we note above, detection is genuinely unreliable, so the candour is welcome, but the gap between what the guidelines admit and what the Code mandates risks eroding confidence in the marking regime just as it takes effect.

The third is the breadth of the exemptions themselves. Some are sensible lower thresholds for the first phase of implementation, but several remain concerning:

minor alterations to input data such as cropping, reframing and AI-generated translation of text;

real-time generation where labeling is technically unfeasible;

text and deepfakes on matters of public interest under other applicable law; and

Several of these are grey areas that market-surveillance authorities will struggle to audit at the scale and speed at which AI content now moves. The personal-use exemption is the most consequential: it lifts the labeling duty from exactly the viral, privately created deepfakes that cause so much harm, pushing the burden downstream onto platforms rather than onto those who make and share them.

Finally, enforcement, oversight and accountability remain open questions, and how these exemptions are read in practice will decide whether the guidelines clarify the obligations or quietly narrow them.

Why is this an important moment for AI regulations across the globe?

Apart from the EU, US states like California and New York are also part of a wave of regulatory efforts focusing on content provenance and authenticity, embedding the same principles that guide initiatives such as the 

Coalition for Content Provenance and Authenticity

 in regulation, and addressing the lack of transparency in AI-generated or manipulated content. We are experiencing a moment where regulators are choosing to act and cover the transparency gaps that have allowed users to be manipulated and trust in what we see and read online to be eroded.

August 2 was an important date, not just due to the provisions of the AI Act going into force, but also as a turning point in AI regulation––from generic transparency measures to focused efforts around remedy and recourse for AI content, including misleading content and misinformation generated by AI systems. That shift will only deliver if transparency is paired with labeling and provenance robust and durable enough to survive how content actually moves, with recourse that is effective for victims and safeguarded against abuse, and with exemptions that do not quietly hand the loudest producers of synthetic media a pass.

Acting now matters because the harm from misuse reaches beyond the people directly targeted. It eats away at public trust, both in these rules and in their promise to help rebuild confidence in what we see and read online.

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

From Tamagotchis to AI companions, generations of machines have been designed to push on our “evolutionary buttons”—ones that ask us to respond to 

In this episode, I spoke with researchers including the Center for Democracy and Technology's 

 to explore the dark patterns in AI companions, and what happens when users go to say “goodbye.”

What follows is a lightly edited transcript of the discussion.

Every day it seems like there's a new headline about people forming attachments with AI chatbots, sometimes with harmful consequences. To learn more, I spoke to Laura Karpas, who's been working with Tech Policy Press this summer as an audio intern from the Craig Newmark Graduate School of Journalism. 

Okay, Justin. So I want to start off with one important question. Do you remember the Tamagotchi?

I do, but I think I am more of the pet rock generation.

So I'm definitely of the Tamagotchi generation. They were this little toy released in 1996, and they were a huge sensation. You remember how they worked, right?

I think more or less you had to press some buttons at a regular interval in order to care for them.

Exactly. They were a handheld video game. They were shaped like an egg. And inside that egg was a digital screen with a digital creature. And to be more specific, it was actually a baby alien. A baby alien that you needed to care for, to water, to feed, to raise it into adulthood. And so if you didn't do any of those things, or if you left it alone… it died.

Definitely compared to a pet rock. And it's important to remember that this type of game where you needed to care for something to respond to a computer's needs, it was really one of the first of its kind. And so people became extremely attached to these little things.

I remember seeing them all over the place for a little while.

Yeah. I mean, people were literally attached to them, Justin. They had a chain that you could hook on your belt. And the thing is, both children and adults, obviously they knew that Tamagotchis were not sentient beings, but it didn't matter. They still wanted to care for and interact with these things. And that's because it tapped into something profoundly human, these Darwinian buttons.

 way back in the '90s, right around the time when the Tamagotchi came out. 

She spoke of this term in the context of human robot relationships.

Latest

Podcast

Recent

Our content delivered to your inbox

Thank you!

Around the globe