transparency

A regulatory package as a long-term political strategy

The European Union’s recent digital laws are often described as a regulatory package. The AI Act, the Data Act, and the emerging Data Union Strategy form a wide experiment in using transparency as infrastructure for the digital economy.

The underlying idea is that digital markets cannot be governed well if users, businesses, regulators, and affected individuals cannot understand how systems work, who controls data, where risks arise, and who is responsible for intervention. Therefore, transparency is becoming a condition for accountability, market access, innovation, and long-term trust that falls under what appears as a long-term strategy to regain data sovereignty.

The EU’s policy bet

The EU regulatory approach is founded on the premise that greater transparency can enhance the governability of complex digital systems. However, the mere disclosure of information does not result directly in a greater understanding of the data available; a company can disclose large amounts of technical material while leaving users no better able to assess risk, compare alternatives, or challenge decisions.

Accordingly, the success of the EU’s transparency framework should not be measured by the sheer volume of regulatory obligations it imposes. Rather, its effectiveness depends on whether those obligations generate information that is genuinely useful in practice. The relevant benchmarks are whether disclosures are meaningful, accessible, timely, and comparable, thereby enabling users and regulators to make informed decisions.

The AI Act’s goal to make AI legible

The AI Act shows the EU’s approach most clearly. Its stated purpose is to improve the functioning of the internal market, promote human-centric and trustworthy AI, protect health, safety, and fundamental rights, and support innovation (Regulation (EU) 2024/1689).

In policy terms, the AI Act tries to make AI systems legible. It assumes that AI risks should not be addressed only after harm occurs. They should be identified, documented, and managed before systems are placed on the market or deployed in sensitive settings.

This is why transparency is linked to risk. High-risk systems face more demanding documentation, monitoring, and information obligations. Lower-risk systems face lighter duties. The European Commission describes the AI Act as the first comprehensive legal framework on AI, designed to address AI risks while fostering trustworthy AI in Europe (European Commission, “Regulatory framework for AI”).

The policy logic is fundamentally pragmatic. Effective regulatory oversight depends on access to adequate information. Likewise, deployers require sufficient information to make informed decisions regarding whether and under what conditions to implement AI systems. Individuals affected by AI-assisted decisions must also have access to relevant information in order to understand how such decisions have been made and, where appropriate, to question or challenge them.

The Data Act attempts to rebalance informational power.

The Data Act uses transparency for a different purpose. Where the AI Act focuses on risk and trust, the Data Act focuses on access, fairness, and economic value. Its objective is to create harmonized rules on fair access to and use of data (Regulation (EU) 2023/2854).

The challenge is that data generated by connected products and digital services is often controlled by a small number of firms. Users may generate valuable data through their use of products but still lack practical access to it. Businesses may need data to innovate, repair products, or offer competing services but face legal, technical, or contractual barriers.

The Commission presents the Data Act as a way to address the challenges and opportunities created by data in the EU, with emphasis on fair access, user rights, and personal data protection (European Commission, “Data Act”).

In this context, transparency functions as a mechanism for redistributing information. Where users are unaware of what data is generated, how it can be accessed, or the conditions under which it may be shared, formally recognized rights of access are unlikely to translate into meaningful practical control. Effective data rights therefore depend not only on their legal recognition but also on the transparency necessary to enable individuals to exercise them.

The Data Union Strategy: From Control to Usable Data

The Data Union Strategy shows the broader direction of EU policy. The Commission frames it around increasing the availability of data for AI development, simplifying EU data rules and strengthening Europe’s position on international data flows (European Commission, “European Data Union Strategy”).

This is significant because it seems that the European Union seeks to pursue two complementary goals simultaneously. On the one hand, it aims to protect fundamental rights and mitigate the risks associated with digital technologies. On the other, it seeks to facilitate greater access to data in order to foster innovation, support the development of artificial intelligence, and enhance European competitiveness. In this way, transparency serves as the connecting principle between these objectives. In fact, by increasing the visibility of how data is collected, processed, and shared, it is intended to strengthen trust in data flows while making them more accessible and capable of supporting innovation.

Why meaningfulness matters most

Meaningfulness is the anchor test. Transparency is useful only if it reveals something that can change decisions or enable scrutiny.

In the AI context, this means information about a system’s purpose, limitations, performance, and risk profile must be specific enough to support procurement, oversight, and challenge. In the data context, it means users must receive information that helps them understand what data exists and how it can be used.

Generic compliance language is not enough. A disclosure that says a system is “risk managed” or that data is “available upon request” may be formally correct but still unhelpful. The real question is whether the information helps someone act.

Information must arrive before decisions are locked in.

Transparency is most useful when it arrives early enough to affect decisions. AI information matters most before procurement and deployment. Data-access information matters most before users become dependent on a particular product, service, or cloud provider.

Post-event transparency can still support audit and enforcement. But it is weaker as a prevention tool. A regime that informs users only after they have lost practical freedom of choice will have limited effect.

Accordingly, comparability occupies a central role in the European Union’s internal market strategy. If transparency is intended to promote competition, facilitate public procurement, and strengthen trust in cross-border digital markets, disclosures must be presented in a manner that enables users, businesses, and regulators to meaningfully compare systems, services, and contractual arrangements.

This objective is particularly relevant in the context of AI procurement, connected product ecosystems, and cloud switching, where informed comparisons are essential to reducing information asymmetries and preventing vendor lock-in. Nevertheless, pursuing comparability inevitably involves trade-offs. While standardized disclosure frameworks can improve the accessibility and consistency of information, they may also obscure sector-specific risks and contextual nuances. Consequently, a uniform template may enhance market discipline and regulatory oversight while simultaneously limiting a more nuanced understanding of the particular risks associated with individual technologies or markets.

The risk of regulatory complexity

The EU’s approach is ambitious, but it is also complex. The AI Act does not operate alone. It sits alongside the GDPR, the Data Act, the Digital Services Act, the Digital Markets Act, the Cyber Resilience Act, and sector-specific rules.

A European Parliament study notes that the AI Act interacts with other digital laws, including the GDPR, Data Act, and Cyber Resilience Act, and that this interplay creates significant regulatory complexity (European Parliament, “Interplay between the AI Act and the EU digital legislative framework”).

Secondary analysis makes a similar point. CEPS has argued that the AI Act may overlap with several horizontal and sector-specific rules, creating possible gaps, inconsistencies, and legal uncertainty (CEPS, “The AI Act and emerging EU digital acquis”).

Competitiveness and the SME problem

The burden of complexity is not shared equally. Large technology firms are better able to absorb compliance costs, hire specialists, and shape standards. Smaller firms may struggle.

Bruegel has warned that EU AI regulation risks imposing disproportionate burdens on smaller firms and may contribute to market concentration if compliance demands are not properly balanced (Bruegel, “The right balance: how to fix European Union artificial intelligence regulation”). This is a key policy tension. The EU wants trustworthy digital markets, but it also wants innovation and technological sovereignty. Transparency can support both goals, but only if it is designed in a way that smaller firms can use and implement.

From disclosure to governance

The EU’s digital strategy should be judged by a practical standard. The question is not whether Europe has created the world’s most elaborate digital rulebook. The question is whether that rulebook produces usable knowledge, enables timely intervention, supports meaningful comparison and redistributes informational power.

If it does, transparency may become genuine governance infrastructure. If it does not, the EU risks building a sophisticated compliance architecture that documents the digital economy without effectively governing it.

NEW YORK — The sudden death of Sen. Lindsey Graham, a top ally of President Trump and one of Washington’s best-known politicians, is renewing focus on the country’s aging lawmakers.

Graham, a South Carolina Republican who had turned 71 just two days before dying on Saturday, was far younger than many of his Senate colleagues and appeared to have been in good health. He suffered a tear in his aorta, according to a preliminary report from the medical examiner.

It was the second time in less than a month that emergency personnel were dispatched to the home of a U.S. senator. In early June, Mitch McConnell of Kentucky, the former Republican Senate leader, was hospitalized for undisclosed reasons.

After weeks of increasingly dire speculation about his health, he finally revealed on Sunday that he had fallen and suffered from mild pneumonia. He released a photo, complete with a copy of the day’s newspaper.

Graham’s death and McConnell’s hospitalization have come amid an ongoing reckoning about the nation’s aging leaders, two years after the disastrous presidential debate that sparked widespread panic among Democrats about then-81-year-old President Biden’s capacities and accusations of a cover-up.

Some politicians have continued to obscure details about their health challenges, asking for privacy despite their public positions, and fueling conspiracy theories.

“I think we need some transparency,” Sen. John Cornyn (R-Texas) said Monday. “I wish Sen. McConnell and his team would have done that earlier. I think it would have resolved a lot of questions.”

McConnell is admitted to a hospital

McConnell, who at 84 is only the third-oldest member of the Senate, was admitted to the hospital on June 14 with barely any explanation. Aides said he was “receiving excellent care” but offered no details about his condition.

The dearth of information fueled a wave of speculation about his prognosis, with Laura Loomer, a Trump ally and conspiracy theorist, claiming on social media that a “high level source close to the White House” had told her he was “officially brain dead.”

But McConnell, who will retire from Congress at the end of January after serving as the longest-ever Senate leader, said in a statement that he is on the mend. He said a fall had led to his hospitalization and that he was “briefly unconscious” and treated for mild pneumonia.

“You all know how folks of my generation often hesitate to share the vulnerability that comes with growing older,” he said. “Even in the public eye, I feel that same instinct — I can’t help it.”

That wasn’t enough to put speculation to rest. On social media, many refused to believe the veracity of a photo his office released that included the front page of the sports section of the Washington Post.

Conspiracy theories about McConnell’s health are “a symptom of our times,” said Sen. Rand Paul, a Republican who is also from McConnell’s home state of Kentucky. Paul said people should “give him a break.”

“People think they have a right to know everyone’s medical problems,” he said, “but I don’t know, where does it begin and where does it end?”

Trump’s medical reports offer limited details

The oldest person ever elected president, at age 78, has long offered only the rosiest picture of his health.

“Everything checked out PERFECTLY,” he boasted after his last physical in May, adding that he took yet another cognitive test aimed at detecting early dementia and has “aced them all.”

His past medical reports have been criticized for offering limited detail and including statistics that some health professionals have viewed with skepticism.

When he first ran for president in 2016, Trump declined to release his health records, breaking with longtime precedent. He instead offered a four-paragraph note from his doctor declaring that he would be “the healthiest individual ever elected to the presidency.” Rep. Ronny Jackson (R-Texas), White House doctor during Trump’s first term, later drew headlines when he extolled the president’s “incredibly good genes.”

When he was infected with COVID-19 in the midst of his 2020 reelection campaign, Trump’s doctors and aides withheld key details of his treatment and tried to downplay the severity of his illness.

And after an attempted assassination at a Pennsylvania rally, Trump aides kept the public in the dark for days, declining to discuss the extent of his injuries or release medical records after assuring he was “fine.”

Kean Jr. goes absent for months

The obfuscation extends beyond the septuagenarian and octogenarian set. New Jersey Republican Rep. Tom Kean Jr. spent four months missing without explanation before he finally disclosed late last month that he had been in treatment for depression.

He said in a brief floor speech after his return that he had remained silent about his condition because he is a “private person by nature.”

He won an uncontested primary during his absence, despite missing more than 100 votes in the House, and is running for reelection.

The approach stood in contrast to Sen. John Fetterman, a Pennsylvania Democrat, who disclosed his hospitalization for clinical depression the day after he was admitted to Walter Reed National Military Medical Center for treatment. He also suffered a stroke while running for office.

Biden’s stumbles doom his reelection effort

Biden’s halting gait, frail appearance and frequent verbal stumbles eventually doomed his 2024 reelection campaign. After a debate in which he frequently lost his train of thought, he chose to withdraw from the race, sparking an unprecedented swap at the top of the Democratic ticket that ultimately paved the way for Trump’s return to office.

Many others have refused to retire. California Sen. Dianne Feinstein, a Democrat, died in office in 2023 at the age of 90, after years of declining health, including a bout of shingles. Though she returned to the Senate after her illness, she appeared frail and confused at times. It was later revealed that her office had failed to disclose in real time that she had contracted encephalitis while recovering.

Longtime Republican Rep. Kay Granger of Texas spent the final months of her more than two decades in Congress, when she was in her early 80s, suffering from what her office called “unforeseen health challenges” that made travel to Washington difficult.

Eleanor Holmes Norton, 89, the longtime House delegate for the District of Columbia, announced earlier this year that she would not run for reelection amid questions about her competency.

Colvin writes for the Associated Press. AP writers Mary Clare Jalonick and Lisa Mascaro in Washington contributed to this report.

For as long as I’ve been a journalist, which is a really long time, public entities have hated public records requests, even while claiming they don’t.

Ask your typical elected or hired official, from the governor to the animal control folks, and they’ll tell you transparency is vital and sunshine in government a key value.

Then turn in the most benign of public records requests — access to a calendar, for example — and prepare for weeks of delays and excuses. Want emails or financial records or, heaven forbid, anything from the police? Months or even years may pass before a single page is delivered, no joke.

That’s why I am deeply concerned about a bill winding its way through the California Legislature that would definitely slow down public records requests and likely make them more difficult and expensive. At its worst, it could push people into costly court battles just for having the audacity to ask for information.

The legislation, Assembly Bill 1821, is authored by Democratic Assemblymember Blanca Pacheco, whose district includes Norwalk, Downey and Bell, where legendary scandals are Example 1 of why public records matter.

Pacheco’s office told me Wednesday that the troubles with the bill are far from what Pacheco set out to do.

“It was never the author’s intention to take away people’s rights to a [Public Records Act] request,” said her chief of staff, Nikki Johnson.

Johnson said the bill was meant to curtail malicious records requests, which do happen, where a citizen goes after copious amounts of records just to be a jerk and cost the government time and money.

It was also meant to address the growing problem of artificial intelligence and other for-profit businesses requesting thousands of records with the intent of using the information to create money-making products — think of sites that already sell publicly available personal information as “background checks.”

I believe Johnson on the good intentions of the bill in addressing those real if nebulous difficulties, but you know what they say about the best-laid plans.

The bill passed through the Assembly recently with ease, largely because most of its problematic portions (I’ll get to those in a minute) were removed — though not all. Even in a watered-down form, which basically gave government more time to answer requests, I found myself in the unlikely position of agreeing with conservative Republican Assemblymember and Trump supporter Carl DeMaio of San Diego, who offered some of the only opposition from elected leaders during the Assembly vote.

“We cannot police the public’s right to know, and we want to err on the side of transparency in how government agencies operate,” DeMaio said.

Amen, brother.

But the Democratic-controlled Assembly erred on the side of secrecy and slowdown instead, and the measure sailed to the Senate, where seemingly out of the blue, a bunch of new provisions were added that fill it with loopholes, vague language and tons of room for abuse.

David Snyder, executive director of the First Amendment Coalition, said the bill as written now was “comprehensively bad for transparency and therefore for government accountability.”

Sean McMorris, transparency, ethics and accountability program manager for the advocacy organization California Common Cause, put it even more forcefully. He pointed out that “public records are the public’s records.”

“They’re not owned by the government,” he said. But this bill would shift that paradigm and make the public “prove why you need them.”

“It’s going to chill people who want to make requests, and it’s going to complicate the process, and it’s just wrong,” McMorris said.

In its new form, the bill basically allows government entities to decide if they feel a public records request is malicious or for commercial gain. If they do, they can petition a court to intervene — potentially sparking both legal costs and new fees associated with fulfilling the request.

It would also, Snyder said, force a requester to explain why they wanted the records — something California law has repeatedly avoided because it gives power to government to treat those it perceives as enemies differently.

In this age of fairness and reason, it’s hard to imagine a government official misusing power to keep secrets, but I’m told it happens. That makes it all the more crucial that people not be forced to explain why they want information, or if they will use it to, say, expose corruption — be it wrongdoing by a single individual or the entire system.

Faced with unintended consequences, Assemblymember Blanca Pacheco (D-Downey), shown in 2023, will seek to scale back the bill to its original form, according to her chief of staff.

(Rich Polk / Getty Images for Equality California)

“I have little doubt that some agencies will use that provision to overburden requesters that they view as political opponents, requesters that they view as just a hassle, requesters that ask for things the government doesn’t want to disclose,” Snyder said. “They can bring the requester into court, and at a minimum, slow down the process, and probably more likely get the requester to simply withdraw.”

As written, the bill also gives a shoddy carve-out meant to protect journalists, but which in reality could be used to curtail requests from freelancers, student journalists and more.

McMorris said access to public records is a “moral issue,” and fixing any problems with the current law requires “a scalpel, not a meat ax.”

This bill, he warned, is a meat ax.

“I don’t discount that there are abusive requests, and that there are requests that really are a burden on government agencies, but the law right now has ways for government agencies to address that,” he pointed out. “Once these laws go into place, they’re going to be hard to roll back.”

It could “fundamentally change” our access to public records, he said.

Johnson, Pacheco’s chief of staff, told me that faced with all these unintended consequences, the Assembly member is going to ask for the amendments to be removed, and for the bill to progress as it was written when it passed the Assembly. That could happen as early as next week, when the bill with the new provisions is scheduled to come up again in a Senate committee for debate.

Reverting to the bill the Assembly voted on would be better, but slowing down public records is in government’s best interests, not the people’s. The bill does nothing to address the problems it seeks to fix, but stretches out the time officials have to simply tell a requester if any records do exist — never mind delivering them.

So even back to its watered-down form, the bill remains a meat ax for a scalpel problem, chopping up transparency with good intentions.