The post Just Security’s Artificial Intelligence Archive appeared first on Just Security.

To achieve that vision, the Plan is centered around three pillars: innovation, infrastructure, and international diplomacy and security. It calls for upskilling the workforce, revising federal rules, building high-security data centers, and tightening export controls—all whilst removing what the administration views as regulatory obstacles to faster AI adoption.

The plan also raises major questions. What’s the role of government in steering this technology responsibly? Are we building the right guardrails as we scale up? And what message is the U.S. sending to allies and adversaries as it charts a new course in AI policy?

Show Notes:

• Sam Winter-Levy’s article, “Assessing the Trump Administration’s AI Action Plan” (July 25, 2025)
• Joshua Geltzer’s article, “The Trump Administration’s AI Action Plan Is Coming. Here’s What to Look For.” (July 18, 2025)
• Alasdair Phillips-Robins’s and Sam Winter-Levy’s article, “What Comes Next After Trump’s AI Deals in the Gulf” (June 4, 2025)
• Clara Apt and Brianna Rosen’s article “Shaping the AI Action Plan: Responses to the White House’s Request for Information” (Mar. 18, 2025)
• Brianna Rosen interview with Joshua Geltzer, “The Just Security Podcast: Trump’s AI Strategy Takes Shape” (April 17, 2025)
• Just Security’s Tech Policy Under Trump 2.0 series
• Just Security’s Artificial Intelligence Archive

The post Decoding Trump’s AI Playbook: The AI Action Plan and What Comes Next appeared first on Just Security.

Even as Trump officials project confidence, the intelligence picture remains fluid and contested. A leaked preliminary report by the Pentagon’s Defense Intelligence Agency (DIA) assessed the operation set Iran’s nuclear program back by mere months, according to several media reports, while CIA Director John Ratcliffe and Director of National Intelligence Tulsi Gabbard have circulated statements suggesting more severe damage, citing “new intelligence” reporting. Evolving intelligence assessments are not unusual in the immediate aftermath of a military operation, as full battle damage assessments typically take weeks to produce, and the preliminary DIA report—made with “low confidence” within 24 hours of the strike—was not coordinated with the broader U.S. intelligence community. 

What is unusual, and deeply troubling, is the overt politicization of the intelligence process on a critical national security issue such as Iran. The Trump administration’s rapid efforts to discredit and sideline internal dissent, delay or restrict intelligence sharing with congressional oversight committees, and elevate favorable Israeli intelligence assessments signals a dangerous shift toward politicization. In a moment of heightened regional instability, the erosion of objectivity within the intelligence community undermines the integrity of U.S. decision-making on Iran and puts broader national security interests at risk. 

What We Know So Far  

According to the New York Times, the leaked DIA assessment concluded the strikes delayed Iran’s nuclear program by fewer than six months, citing evidence that the strikes sealed off the entrance to two of the nuclear sites but did not collapse their underground buildings. The assessment also reportedly claimed Iran moved much of its uranium enrichment stockpile prior to the strike, suggesting it retains access to highly enriched nuclear material. Notably, the text of the leaked assessment has not been made public yet, and Trump’s personal attorney has disputed and threatened to sue the New York Times for libel over its characterization of the DIA assessment. 

The CIA and Office of the Director of National Intelligence (ODNI) have presented a starker picture, although one that is potentially not inconsistent with the initial DIA report. In a statement, Ratcliffe confirmed the CIA had new intelligence from a “historically reliable” source that “several key Iranian nuclear facilities were destroyed and would have to be rebuilt over the course of years.” It is not clear yet what he meant by “several facilities” and whether these are within the three nuclear sites that were struck. Gabbard similarly asserted new intelligence reporting indicates Fordow, Natanz, and Esfahan were “severely damaged” and “would likely take years to rebuild.” Gabbard previously testified before Congress in March that Iran was not building a nuclear weapon, but seemed to walk back that line after Trump publicly criticized her assessment as “wrong.”  

Israeli and Iranian officials, for their part, have offered mixed damage assessments. On June 25, Israeli Defense Forces (IDF) Chief of Staff Gen. Eyal Zamir asserted Iran’s nuclear program had suffered “severe, extensive, and deep damage and has been set back by years.” In the first pre-recorded speech since the ceasefire between Israel and Iran, Iranian Supreme Leader Ayatollah Ali Khamenei claimed victory over the United States and Israel, echoing his earlier statement that Trump’s assessment of the damage to Iran’s nuclear program was “exaggerated.” The Supreme Leader’s remarks appeared in tension with previous statements from the Iranian Foreign Ministry spokesperson, who said Iran’s nuclear installations were “badly damaged” in the U.S. and Israeli strikes.

Even if the United States did severely damage the three nuclear sites, it does not necessarily follow that they set Iran’s overall nuclear program back by years. It may indeed take Iran years to rebuild those facilities but, crucially, Iran does not need to rebuild all three to weaponize. Some Israeli officials assess Iran maintains small covert enrichment facilities that would allow it to continue its nuclear program in the event of an attack on its larger facilities, raising the prospect of Iran being able to construct a crude nuclear device. Striking nuclear facilities, or even nuclear scientists themselves, does not eliminate the knowledge Iran has already acquired of the weaponization process. 

Irrespective of what the damage assessment ultimately says, there is no firm evidence yet that Iran’s nuclear program was destroyed. Gathering such evidence will take time, requiring monitoring from the International Atomic Energy Agency (IAEA), should Iran allow inspectors back in, or human intelligence sources of the ground. On Thursday, the IAEA Director General Rafael Grossi suggested it would be difficult to evaluate damage from the strikes based on satellite imagery alone, while confirming that centrifuges at Fordow were “no longer operational.” Preliminary intelligence assessments reportedly provided to European governments indicate Iran’s enriched uranium stockpile remains largely intact. 

In the coming weeks, additional technical assessments will follow, with Ratcliffe pledging to “provide updates and information to the American people.” Even then, the ultimate effectiveness of the strikes will remain a matter of debate, as Iran could maintain covert enrichment facilities or uranium stockpiles that were either unknown to U.S. intelligence or moved before or after the strikes. 

The U.S. intelligence community now faces the difficult task of parsing through satellite imagery, signals intelligence, human intelligence, allied intelligence reports, and open source information to try to present policymakers with the fullest picture of the facts. This is never a straightforward process, as even Iranian decisionmakers may not yet understand the full extent of the damage, or might be misled by their own military officials seeking to downplay it. The risk of mis- and disinformation, as well as covert deception campaigns, is high. In the days ahead, the intelligence community must remain laser focused on presenting the facts with analytic integrity, a task made immeasurably more difficult by the increasing politicization of intelligence under Trump. 

Growing Signs of Politicization 

On Thursday, the Senate received a classified briefing on the strikes, with a briefing for the House planned for tomorrow. Ratcliffe represented the intelligence community, alongside Hegseth, Caine, and Secretary of State Marco Rubio, with Gabbard notably absent. Several Democrats leaving the briefing expressed skepticism about the Trump administration’s claims, with Senator Chris Murphy (D-CT) asserting it “still appears we have only set back the Iranian nuclear program by a handful of months.” The White House previously postponed classified briefings on the operation required under the 1973 War Powers Resolution, prompting alarm among lawmakers. “I’m very concerned about [Trump] distorting, manipulating, and even lying about intelligence,” Sen. Chris Van Hollen (D-MD) said. “We’ve been here before, We went to war in Iraq under false pretenses.” 

The Trump administration has said it plans to limit intelligence sharing with Congress following the unauthorized disclosure of the DIA report, which White House Press Secretary Karoline Leavitt characterized as “flat-out wrong” and leaked by a “low-level loser in the intelligence community.” The FBI has since opened an investigation into the source of the leaks. 

In an unprecedented move, the Trump administration has promoted Israeli intelligence assessments and even selective Iranian statements to bolster its case on the effectiveness of the strikes. A White House statement entitled, “Iran’s Nuclear Facilities Have Been Obliterated — And Suggestions Otherwise are Fake News” cites the IDF Chief of Staff and Iranian Foreign Ministry Spokesperson, alongside U.S. officials. 

The Iran strikes are not the first indication of growing politicization within the intelligence community. In May, Gabbard fired two senior officials at the National Intelligence Council (NIC) after the group produced a partially declassified intelligence assessment contradicting Trump’s rationale for invoking the Alien Enemies Act to deport alleged Venezuelan gang members. In emails obtained by the Times from DNI Chief of Staff Joe Kent to senior ODNI officials, Kent pushed for “some rewriting” so the NIC assessment “would not be used against the DNI or POTUS [President Trump].” That is after he had earlier pressed analysts to redo a Feb. 26 assessment of the Venezuelan matter. “Some intelligence officials took Mr. Kent’s intervention as an attempt to politicize the findings and push them in line with the Justice Department arguments and the Trump administration policy,” the Times reported.

More broadly, Gabbard has sought to consolidate control over the intelligence process by moving the NIC to ODNI headquarters and attempting to do the same with the office responsible for compiling the President’s Daily Brief (PDB). Since Trump returned to power, senior intelligence personnel have been removed or reassigned as part of a wider restructuring of the intelligence community. 

What Congress Should Do Now 

The increasing politicization of the intelligence community on critical national security matters puts everyone at risk. But Congress is not powerless to act. Lawmakers can take a number of steps in the immediate and longer term to ensure that intelligence remains politically neutral. 

First, Congress must press for the fullest possible picture of what happened during Operation Midnight Hammer. That begins with robust, closed-door briefings that feature career analysts alongside senior appointees, so lawmakers can hear the range of professional judgments and interrogate assumptions. 

In the longer term, a careful, bipartisan review—akin to the Iraq or Afghanistan War Commision but on a smaller scale—would clarify the facts surrounding this strike and send a signal that intelligence must not be politicized as a matter of national security. As part of such a review, Congress should scrutinize the unusual prominence accorded to Israeli intelligence assessments to ensure that U.S. intelligence is not being sidelined. Allies provide invaluable insight, yet each has its own strategic calculus in doing so. Elevating any external narrative above the collective judgment of the U.S. intelligence community risks outsourcing American decision-making to Israel. The very creation of such a body can help insulate the intelligence community from political interference going forward. It would also send an important signal that, on this and other issues, there is a prospect of external review that can detect and deal with any politicization of intelligence work. 

Alongside these actions, broader measures are urgently needed within the intelligence community to protect whistleblowers, safeguard independent analysis, and ensure that leak investigations are not used as instruments of intimidation against dissenting analysis or inconvenient assessments. If career analysts believe they will face retribution for assessments that do not align with White House policy, the entire intelligence and decision-making process collapses. One step, as former senior CIA executive Brian O’Neill has proposed in these pages, would be for Congress to establish a standing analytic review board—independent of the ODNI—to adjudicate internal complaints about intelligence manipulation, coercion, or suppression. Other potential steps include establishing an analytic integrity review panel, modeled on the State Department’s Historical Advisory Committee, which would serve as a neutral, public-facing check on institutional drift. Congress could require quarterly testimony from this review panel, as well as from the ODNI’s Analytic Ombuds Office and senior officials responsible for producing the PDB, to ensure political pressure or bias are not shaping intelligence analysis. 

Finally, Congress should push the Trump administration to be transparent with the American people about the effectiveness of the strikes without delay. The administration can do this by declassifying intelligence assessments to the fullest extent possible, while protecting sources and methods. 

By asserting rigorous oversight now, Congress can begin to repair the damage being done to the intelligence community. The goal of such oversight should not be to score shortsighted partisan victories, but rather to ensure that any future course of action on Iran and other matters is anchored in accurate, robust intelligence. Without that foundation, even the most decisive tactical successes ultimately risk becoming strategic liabilities. 

The post Intelligence Implications of the Shifting Iran Strike Narrative appeared first on Just Security.

These developments raise urgent questions about the risks of regional escalation, the legality of unilateral military action, and the long-term implications for the global nonproliferation regime. The following analysis aims to assist policymakers, lawmakers, journalists, and the public assess the implications of this moment. For further expert commentary, see other installments in Just Security’s collection on the Israel-Iran conflict. 

Iran’s Retaliatory Options

Iran now faces a range of constrained but potentially destabilizing choices. Iranian leaders likely will try to reassert deterrence through a show of continued strength, while avoiding actions that could trigger full-scale war with the United States. Setting aside legal considerations, the below options are not mutually exclusive and carry varying levels of risk in terms of conflict escalation and potential blowback to Iran: 

Missile and Drone Strikes. Iran maintains a diverse and advanced missile stockpile, with some capable of targeting the United States and its allies, although the Israeli government claims to have destroyed half of Iran’s missile launchers since June 13. Tehran may opt for a calibrated strike—similar to its 2020 missile attack on Al-Asad Airbase in Iraq, which houses U.S. and allied forces, following the U.S. drone strike that killed Iranian general Qassem Soleimani in Iraq. (That strike resulted in over 100 U.S. troops suffering traumatic brain injuries.) Such a strike could demonstrate resolve while aiming to stay below the threshold of major escalation, but it may not be viewed within Iranian leadership as a sufficient response to the U.S. use of force against Iran. 

Proxy and Terrorist Attacks. Iran may activate regional proxies such as Hezbollah in Lebanon, Shia militias in Iraq and Syria, or the Houthis in Yemen. These groups could target U.S. personnel, military installations, or allied interests in the region. The Houthis have already threatened to target U.S. naval vessels in the Red Sea in response to the U.S. strikes. A Hezbollah spokesperson, by contrast, indicated the group has no immediate plans to retaliate following the U.S. strikes. 

Even if these groups do join the fighting, Iran’s most capable regional proxies—Hezbollah and Hamas—have suffered major setbacks since the Israeli response to the Hamas-led attacks of Oct. 7, 2023. Israeli officials estimate that Hezbollah maintains only some 20 percent of the missiles and rockets it had before the war. Dozens of senior Hezbollah commanders, including the longtime leader Hassan Nasrallah, have been killed in targeted Israeli strikes. Hamas, meanwhile, has seen much of its leadership killed or displaced, and its military infrastructure in Gaza severely degraded by sustained Israeli operations.

Tehran may also turn to transnational terrorism. The direct use of force against Iran could prompt attacks beyond the region. Hezbollah-linked operatives and Iranian sleeper cells may pose a potential threat to Israeli, European, or U.S. interests. Iran has demonstrated the intent and capacity to orchestrate extraterritorial attacks, including a foiled 2011 attempt to assassinate the Saudi ambassador in Washington, DC; a 2022 murder-for-hire plot targeting Iranian-American journalist Masih Alinejad in New York City; alleged plots against former National Security Advisor John Bolton and former Secretary of State Mike Pompeo; and reported threats against Trump in 2024. 

Maritime Disruption. Iran could attempt to disrupt maritime traffic in the Strait of Hormuz through mining, drone activity, or harassing commercial and military vessels. Such action would threaten global energy markets and signal Tehran’s willingness to retaliate, but also pose a high risk of direct confrontation with U.S. and allied forces that could threaten regime survival. Such a move also risks angering China—the largest importer of Iranian oil—whose economy is dependent on energy transported through the Strait, and has so far remained relatively neutral in the conflict. 

Cyber Operations. Cyber retaliation remains a likely and strategically flexible option. Iran has previously targeted critical infrastructure through cyberattacks and has intensified such operations amid ongoing conflict with Israel. Iranian malicious cyber activity reportedly surged by more than 700 percent since June 13. 

A cyber response could allow Iran to impose costs while retaining plausible deniability. Potential targets include energy infrastructure, telecommunications systems, financial networks, and other critical infrastructure within the United States and abroad. U.S. companies are on high alert, but defensive gaps remain—particularly in light of the Trump administration’s recent staff reductions and proposed additional budget cuts at the U.S. Cybersecurity and Infrastructure Security Agency. 

Cyber escalation is inherently unpredictable. Even without deliberate targeting, malicious code can propagate widely, as seen in the unintended global spread of Stuxnet—an alleged U.S.-Israel cyberweapon designed to disrupt Iran’s centrifuges that ultimately infected companies based in the United States.

Withdrawal from the Nuclear Non-Proliferation Treaty (NPT). Iran may choose to withdraw from the NPT and reconstitute its nuclear program outside international oversight. (The treaty obligates non-nuclear-weapon States to forgo the development or acquisition of nuclear weapons, while allowing access to peaceful nuclear technology under IAEA safeguards, among other constraints.) While this would carry significant diplomatic and economic consequences, it could serve as a form of strategic signaling and raise the specter of a nuclear breakout (see further discussion on potential NPT withdrawal below.)

Legal Status of the Strikes

From a domestic legal perspective, the U.S. Constitution vests the power to declare war in Congress. The June 21 strikes were carried out without congressional authorization, and the Trump administration has not invoked any existing statutory basis—such as the 2001 or 2002 Authorizations for the Use of Military Force (AUMFs)—to justify the operation. As Brian Egan and one of us (Tess Bridgeman) have explained, neither of those two AUMFs authorize war against Iran; moreover, direct strikes against Iran on its territory with the obvious threat of regional escalation almost certainly rise to the level of “war in the constitutional sense,” such that the President could not rely on Article II authority alone but would need congressional authorization. 

The War Powers Resolution requires the president to notify Congress within 48 hours of hostilities and requires withdrawal from otherwise unauthorized military operations after 60 days unless Congress provides authorization. Congress could also try to pass a bill under the War Powers Resolution to require withdrawal immediately (a few such bills have already been introduced), but it would have to overcome a potential presidential veto to be enacted. Regardless, the absence of any legislative authorization in such a consequential use of force raises serious concerns about democratic accountability and the erosion of Congress’ constitutional role.

From an international legal perspective, the U.N. Charter prohibits the use of force against the territorial integrity of another State under Article 2(4), except in narrow cases of self-defense under Article 51 or with U.N. Security Council authorization. The United States has not yet publicly presented evidence of an imminent armed attack that would justify self-defense under Article 51. 

The United States also has not cited a formal request for collective self-defense from Israel, although Israel has publicly called on the United States to do so. At a press conference following the U.S. strikes, Secretary of Defense Pete Hegseth said, “the president authorized a precision operation to neutralize the threats to our national interests posed by the Iranian nuclear program and [in support of] the collective self-defense of our troops and our ally, Israel.” Any formal invocation of collective self-defense, however, would merit close scrutiny, considering the legality of Israel’s military intervention is subject to debate (see here and here). In the absence of either a self-defense or collective self-defense justification, the strikes may be viewed as preventive, placing them on uncertain legal footing at best. 

Even if the United States does present such a case, any use of force under international law must meet the requirements of necessity and proportionality. The United States has not yet explained how targeting multiple nuclear facilities within a sovereign country—absent an imminent threat originating from those sites—comports with these legal requirements. Importantly, the imminent threat would need to encompass not only Iran’s potential to acquire a nuclear weapon soon but also an intention to use it. As Adil Haque recently wrote, “Any notion that Iran intends to trigger mutually assured destruction is delusional and cannot provide a rational basis for the use of force under international law.”

One key question is if President Trump is correct that “Iran’s key nuclear enrichment facilities have been completely and totally obliterated,” then the U.S. and Israeli operations would have achieved their objective. In such a case, the legal basis for the continued use of force would no longer be necessary or proportionate.

(Note: we do not address here the rules of armed conflict that apply to attacks on nuclear facilities.)

Implications for the Nonproliferation Regime

On June 12, the International Atomic Energy Agency’s Board of Governors found Iran in “non-compliance” with its safeguards obligations, marking the second such finding in two decades. Notably, Iran was in compliance with its NPT obligations and additional monitoring measures under the 2015 Joint Comprehensive Plan of Action (JCPOA) until the United States withdrew from the deal in 2018. However, Iran still remains a party to the NPT, which legally requires it to permit IAEA inspections of its declared nuclear facilities.

The U.S. strikes place IAEA verification and monitoring at serious risk and may have significant consequences for the credibility of the NPT, which is regarded as the cornerstone of the global nuclear nonproliferation regime: 

• Undermining the Inspection Process. While the IAEA has formally declared Iran to be in breach of its nonproliferation obligations and has decreased international access since the U.S. JCPOA withdrawal, attacking a State under IAEA monitoring, particularly before exhausting diplomatic options, may still be perceived as punishing transparency and cooperation. This risks discouraging other States from voluntarily submitting to IAEA safeguards out of concern that openness could increase their vulnerability to military action. 
• Risk of NPT Withdrawal. Iran could choose to invoke Article X of the NPT and withdraw from the treaty, following North Korea’s path in 2003. Taking this step could allow Iran to reconstitute and expand its nuclear program outside most legal or technical constraints.
• Precedent for Preventive Force. The strikes may set a corrosive precedent for the use of force against nuclear infrastructure in the absence of clear evidence of an imminent threat. This challenges the core logic of the NPT—that proliferation concerns should be addressed through verification and diplomacy, rather than unilateral military action.

More broadly, these developments could reinforce a perception that non-nuclear-weapon States remain vulnerable to coercion or attack, while nuclear-weapon States may act with impunity. Over time, such perceptions may erode trust in the nonproliferation regime and incentivize other States to pursue nuclear capabilities as a means of deterrence.  

On the other hand, an argument sometimes raised is that the use of force can reinforce the nonproliferation regime by deterring States from developing nuclear weapons or violating their legal obligations under the NPT. Yet this line of argument would be widely understood by foreign capitals as a rejection of the diplomatic choices in constructing the NPT itself. It is difficult to see how such an approach and the NPT could both remain viable.  

The post The Day After U.S. Strikes on Iran’s Nuclear Program: A Policy and Legal Assessment appeared first on Just Security.

How do leaders steer through cyber crises, build trust, and chart a path forward?

In conversation with Dr. Brianna Rosen, Just Security Senior Fellow and Director of the AI and Emerging Technologies Initiative, Jen Easterly, who just completed a transformative tenure as Director of CISA under the Biden Administration, unpacks the challenges, breakthroughs, and lessons from the front lines of America’s cybersecurity efforts.

Show Notes: 

• Jen Easterly
• Just Security’s Cybersecurity coverage
• Empathy Matters: Leadership in Cyber by Jen Easterly ( 2019)

The post The Just Security Podcast: A Conversation with Jen Easterly — Cybersecurity at a Crossroads appeared first on Just Security.

As western unity frays and U.S. diplomacy shifts under President Trump, the war has become a flashpoint for competing visions of the international order. 

This week, the European Union gave Russia an ultimatum: accept a proposed ceasefire or face expanded sanctions—just days ahead of a potential round of direct peace talks in Istanbul on Thursday. The stakes are high, and the choices made this week could reshape not only the trajectory of the war but the future of global security.

How should we understand the prospects for a sustainable peace in Ukraine amidst evolving geopolitical dynamics and continued battlefield uncertainty? 

To help make sense of these developments, Just Security Senior Fellow and Director of the Oxford Programme for Cyber and Tech Policy, Brianna Rosen, sat down with Sir Lawrence Freedman, Emeritus Professor of War Studies at King’s College London and Professor Janina Dill, Dame Louise Richardson Chair in Global Security at Oxford University’s Blavatnik School of Government. 

This conversation was part of the Calleva-Airey Neave Global Security Seminar Series at the University of Oxford.

Show Notes: 

• Just Security’s Russia-Ukraine War Archive 
• Ambassador Daniel Fried’s “How to Land the Emerging Peace Deal on Peace for Ukraine“
• Music: “Broken” by David Bullard from Uppbeat: https://uppbeat.io/t/david-bullard/broken (License code: OSC7K3LCPSGXISVI)

The post The Just Security Podcast: Peace Diplomacy and the Russo-Ukraine War appeared first on Just Security.

As the Trump administration is poised to unveil its AI Action Plan in the near future, the broader contours of its strategy for AI adoption and acceleration already seem to be falling into place.

Is a distinct Trump strategy for AI beginning to emerge—and what will that mean for the United States and the rest of the world?  

Show Notes:

• Joshua Geltzer
• Brianna Rosen
• Just Security series, Tech Policy Under Trump 2.0
• Clara Apt and Brianna Rosen’s article “Shaping the AI Action Plan: Responses to the White House’s Request for Information” (Mar. 18, 2025)
• Justin Hendrix‘s article “What Just Happened: Trump’s Announcement of the Stargate AI Infrastructure Project” (Jan. 22, 2025)
• Sam Winter-Levy‘s article “The Future of the AI Diffusion Framework” (Jan. 21, 2025)
• Clara Apt and Brianna Rosen’s article, “Unpacking the Biden Administration’s Executive Order on AI Infrastructure” (Jan. 16, 2025)
• Just Security’s Artificial Intelligence Archive
• Music: “Broken” by David Bullard from Uppbeat: https://uppbeat.io/t/david-bullard/broken (License code: OSC7K3LCPSGXISVI)

The post The Just Security Podcast: Trump’s AI Strategy Takes Shape appeared first on Just Security.

In the months since the presidential transition, more than 8,000 webpages and thousands of datasets have been removed or altered across agencies, including the Centers for Disease Control and Prevention, Census Bureau, and Food and Drug Administration. Removal of this data from government agency websites jeopardizes access to vital public information on health, safety, environmental, and demographic issues. At the same time, sweeping executive orders on interagency data-sharing and the consolidation of sensitive information under the Department of Government Efficiency (DOGE) have drawn criticism from privacy advocates, transparency experts, and civil society groups. These concerns are magnified by weakened oversight mechanisms following the dismissal of key members of the Privacy and Civil Liberties Oversight Board. 

In response to these developments, Just Security is launching a new series, “Data Preservation Under the Trump Administration,” to explore the legal, ethical, and policy implications of current federal data practices. As political pressures intensify and the risk of data loss grows, this series aims to provide comprehensive analysis of how federal data is stored, shared, and accessed. 

The series contains the following articles and will be updated regularly: 

• Deborah Brown, “DOGE’s Growing Reach Into Personal Data: What it Means for Human Rights” (Apr. 16, 2025)
• Amy Edwards Homes, “Vanishing Accountability: The U.S. Retreat from Open Data”  (Apr. 17, 2025)
• Corinna Turbes, “Beyond Data Rescue: Building Structural Safeguards for Federal Data Preservation” (Apr. 28, 2025)
• Richard Painter, “Federal Records Belong to the Government” 

IMAGE: Blue U.S. Capitol building in front of a red background of data (via Getty Images)

The post Introduction to Series: Data Preservation Under the Trump Administration appeared first on Just Security.

While diverse in approach, the submissions converge around several core themes: infrastructure and energy development, federal preemption of state AI laws, export controls to maintain U.S. competitiveness against rivals like China, promoting domestic AI adoption, safeguarding national security, and defining clear copyright and licensing frameworks for AI data. What follows is a thematic roundup of these proposals, culminating in a reference table at the end.

Innovation, Workforce Adoption & Economic Impacts

A dominant concern among commenters is how the federal government can accelerate AI growth and create a robust domestic workforce. Many commenters worry that a patchwork of state laws might create a “fragmented regulatory environment.” Google, OpenAI, and the Business Roundtable — an association representing U.S. CEOs — explicitly support federal preemption of state-level AI regulations that create a fragmented regulatory environment which could “undermine America’s innovation lead.”

The News/Media Alliance, CDT, and CSET, meanwhile, call for measures that support “Little Tech” to avoid a market dominated by a few large AI providers. The latter two, in particular, highlight the importance of supporting open-source models, which can enable “greater participation” in the AI domain by lowering barriers to entry for smaller firms with fewer resources.

Additionally, some organizations (Anthropic, Business Roundtable, Google, Center for Data Innovation, MITRE) spotlight the labor market implications of AI, urging the administration to invest in technical education and workforce training. To address labor shortages in high-demand, AI-related jobs, CNAS and Google both propose leveraging immigration authorities, including expediting visa applications. Anthropic also recommends that the White House monitor and report on how AI reshapes the national economy, including its effect on the tax base and labor composition.

Export Controls and Global AI Leadership

Concern over China’s rapid AI progress permeates nearly every submission. OpenAI cites the rise of DeepSeek as an example of the country’s swiftly narrowing AI gap, warning that the Chinese Community Party’s (CCP) subsidized efforts could undermine U.S. advantages. Anthropic and CNAS similarly stresses the urgency of preventing smuggling of advanced chips to China, recommending new government-to-government agreements to close supply chain loopholes. Most of the submissions include calls to strengthen the U.S. Commerce Department’s Bureau of Industry and Security, such as by increasing its funding, instituting scenario planning assessments prior to implementing export controls, and consulting with the private sector.

On the other hand, the Center for Data Innovation warns that current, often “reactive” U.S. export controls hamper U.S. firms’ global competitiveness “without meaningfully slowing China’s progress.” The Center suggests a pivot toward enhancing domestic AI capabilities, streamlining export licensing, and collaborating with allies to promote a democratic vision for AI standards.

Google’s position is that any new rules — particularly the AI Diffusion Rule, set forth by the Biden administration — should avoid imposing “disproportionate burdens on U.S. cloud service providers” and factor in potential adverse impacts on American market share. Google, OpenAI, and Anthropic underscore that effective export controls should avoid inadvertently accelerating foreign AI development.

CDT, CNAS and CSET underscore the strategic importance of the United States to “remain at the frontier” of open-source models in the AI race against China. They argue that without compelling U.S. alternatives,  China could embed “authoritarian values” in AI models adopted by developing countries. To counter this, CNAS proposes that the United States rapidly release modified versions of open-source Chinese models to “strip away hidden censorship mechanisms” and promote democratic values abroad.

Infrastructure and Energy

Submissions from OpenAI, Anthropic, Google, CNAS, and Business Roundtable eachpress for robust infrastructure and energy reforms to meet AI’s skyrocketing computational demands. OpenAI and CNAS propose establishing special zones to attract massive private investment in new data centers and transmission lines, as well as to “minimize barriers” and “eliminate redundancies” through tax incentives, streamlined permitting, and partial exemptions from the National Environmental Policy Act.

Anthropic proposes a national target of “50 additional gigawatts of power dedicated to the AI industry by 2027,” cautioning that if the United States fails to supply reliable, low-cost energy, domestic AI developers might relocate model training to authoritarian countries, exposing U.S. intellectual property to theft or coercion. Google underscores the need for consistent federal and state incentives to promote grid enhancements, data center resilience, and advanced energy-generation projects.

Government Adoption of AI

The majority of submissions highlight lagging AI adoption by federal agencies. OpenAI characterizes government usage as “unacceptably low,” proposing to waive or streamline certain compliance requirements to accelerate pilot programs. Anthropic goes further, calling for a government-wide audit to “systematically identify” every text, image, audio, and video workflow that could be AI-augmented.

Proposals also emphasize removing procurement barriers in both civilian and national security contexts. Anthropic wants to mobilize the U.S. Department of Defense (DoD) and Intelligence Community (IC) to expedite AI adoption, while Google and CSET urge the government to avoid duplicative or siloed AI compliance rules across agencies. The Center for Data Innovation warns against the government’s “risk-only” mindset, imploring the administration to pivot to an “action” framework that proactively integrates AI where it can transform mission delivery. CNAS also advises that the U.S. military take “full advantage” of AI and autonomous systems, provided that the DoD develops “rigorous and streamlined” testing procedures and “permit warfighters an early and ongoing role.”

On the other hand, CDT’s proposal cautions against the government rushing forward on AI adoption, claiming that it could lead to “wasted” tax dollars on ineffective, “snake oil” AI tools. CDT instead advocates for stronger guardrails on government AI usage, including the establishment of an independent external oversight mechanism to monitor AI deployment in national security and intelligence contexts. It further recommends that agencies expand existing use case inventories to transparently catalogue how AI systems are being utilized. Notably, CDT urges that the Trump administration should “clarify and proactively communicate” how the Department of Government Efficiency (DOGE), an unofficial federal body led by Elon Musk, is reportedly using AI to make high-risk decisions.

AI Security and Safety

Anthropic, Google, and the Center for Data Innovation each underscore the national security implications of frontier AI models. Anthropic’s submission notes that new AI systems are trending toward capabilities that could facilitate the development of biological or cyber weapons, emphasizing the need for “rapidly assessing” advanced models for potential misuse. To mitigate such risks, the Center for AI Policy (CAIP) recommends that the U.S. government establish a clear definition of frontier AI so that national security regulations effectively address the most high-risk models.

Anthropic and others also advocate for keeping the U.S. AI Safety Institute (AISI) intact while bolstering it with statutory authorities and interagency coordination to test AI models for national security risks. The Center for Data Innovation, CNAS, and CSET, meanwhile, propose creating a national AI incident database and a vulnerability database — akin to the National Vulnerability Database for cybersecurity — to track AI failures, identify systemic weaknesses, and coordinate risk mitigation. CAIP takes this a step further, urging the Trump administration to create an “AI Emergency Response Preparedness Program” involving “realistic simulations” of AI-driven threats — including AI-enabled drone and cyber attacks — requiring AI developers to respond to these scenarios.

Google, CNAS, and CSET call for collaboration with private labs and the IC to evaluate and mitigate potential security threats, including espionage and chemical, biological, radiological, and nuclear (CBRN) vulnerabilities. Google also opposes mandated disclosures that could reveal trade secrets or model architecture details, warning that such transparency “could provide a roadmap” for malicious actors to circumvent AI guardrails.

CDT  suggests the U.S. AI Action Plan should incorporate measures to address other AI safety risks, including privacy violations and discrimination concerns. It recommends that the National Institute of Standards and Technology should “holistically and accurately” assess the efficacy and fairness of AI systems, as well as issue guidance on evaluating the validity of measurements used. Lastly, CSET proposes that the Trump administration create “standard pathways” to challenge “adverse” AI-enabled decisions and implement whistleblower protections at frontier AI firms to discourage “dangerous” practices.

Obligations for AI Developers, Deployers, and Users

A recurring theme, particularly in Google’s submission, is the need to clearly delineate liability throughout the AI lifecycle. Google argues that developers cannot be held responsible for every downstream deployment — particularly when they lack control or visibility into final uses. Instead, they advocate “role-based” accountability: Developers should provide transparency around model training, but the ultimate deployers should bear liability for misuse in their applications.

At the same time, Google concedes that certain minimal disclosures (for instance, about synthetic media) may be warranted, but it resists broad, mandatory “AI usage” labels that could inadvertently help adversaries “jailbreak” or circumvent AI security features.

Copyright Issues and Development of High-Quality Datasets

OpenAI, Google, the Center for Data Innovation, and MITRE each argue for policies that expand access to robust, high-quality datasets while preserving fair use protections. OpenAI maintains that applying fair use to AI is a “matter of national security” in the face of Chinese competitors who “enjoy unfettered access” to copyrighted data. The company warns that narrowing data access could give Beijing an irreparable advantage in the race to develop state-of-the-art AI models.

The News/Media Alliance, representing over 2,000 media organizations, focuses on publisher rights. It raises concerns that generative AI models are trained on vast quantities of copyrighted material without permission, threatening traditional news revenue streams. The Alliance proposes collaborative licensing agreements and clearer guidelines about disclosing when and how AI-generated content uses news materials.

Finally, the Center for Data Innovation recommends a National Data Foundation, analogous to the National Science Foundation, that would fund the creation, structuring, and curation of large-scale datasets across both public and private sectors. Business Roundtable also highlights the importance of unlocking access to government datasets from the perspective of producing more representative, less biased AI models.

* * *

The table below highlights key aspects of public submissions responding to the White House’s RFI.

IMAGE: Visualization of an AI chip (via Getty Images)

The post Shaping the AI Action Plan: Responses to the White House’s Request for Information appeared first on Just Security.

What are the key takeaways from the Summit? How might it shape other global efforts to regulate artificial intelligence?

Joining the show to discuss the Summit is Dr. Brianna Rosen, Director of Just Security’s AI and Emerging Technologies Initiative and Senior Research Associate at the University of Oxford.

Show Notes:

• Just Security’s Artificial Intelligence coverage
• Just Security’s Tech Policy under Trump 2.0 Series
• Music: “Parisian Dream” by Albert Behar from Uppbeat: https://uppbeat.io/t/albert-behar/parisian-dream (License code: RXLDKOXCM02WX2LL) 

Listen to the episode, with a transcript available, by clicking below.

The post The Just Security Podcast: Key Takeaways from the Paris AI Action Summit appeared first on Just Security.