His announcement that the agreement is contingent upon the kingdom’s recognition of Israel throws the deal’s future into doubt.
September 2026
By Kelsey Davenport
The United States signed a nuclear cooperation agreement with Saudi Arabia, but President Donald Trump’s announcement that the deal is contingent upon the kingdom’s recognition of Israel throws the deal’s future into doubt.
U.S. Energy Secretary Chris Wright and Saudi Energy Minister Prince Abdulaziz bin Salman signed the nuclear cooperation agreement, commonly known as a “123 agreement,” July 20 and July 22, respectively.
The agreement, which is necessary before U.S. companies can sell certain nuclear materials and related technologies, was expected after the White House announced a framework for nuclear cooperation with Saudi Arabia in November 2025. (See ACT, December 2025.)
In a statement issued after the signing, Wright said that the agreement upholds “the highest standards of nuclear safety and nonproliferation.” The deal, however, supports Saudi Arabia’s plan to acquire uranium enrichment technologies necessary to produce nuclear fuel, without insisting on more intrusive International Atomic Energy Agency (IAEA) monitoring. That same technology can be used to produce fissile material for nuclear weapons.
Saudi Arabia does not have a nuclear weapons program and pledged as a member of the nuclear Nonproliferation Treaty (NPT) not to acquire such weapons, but officials, including Crown Prince Mohammed bin Salman, have threatened to build nuclear weapons if Iran does. (See ACT, June 2018.) According to documents obtained by Arms Control Today, Trump formally signed the required presidential determination that the agreement does not “constitute an unreasonable risk to common defense and security” July 16, despite those threats.
The day after Wright signed the deal, Trump posted on Truth Social that the nuclear cooperation agreement “will be approved, but is totally subject to Saudi Arabia joining the very respected and successful Abraham Accords.”
Saudi Arabia has not responded to Trump’s demand regarding the Abraham Accords, a diplomatic framework that led several countries, including the United Arab Emirates and Bahrain, to recognize Israel, but Israeli Prime Minister Benjamin Netanyahu praised the decision. In a July 23 statement, Netanyahu said that Saudi Arabia joining the accords would be a “historic leap forward for peace in the Middle East.”
The Trump administration submitted the nuclear cooperation agreement and the required nonproliferation assessment to Congress for a 90-day review, as required by the Atomic Energy Act on Aug. 24.
The deal is expected to receive pushback from Republican and Democratic legislators, but it is unclear if there will be enough opposition to pass a joint resolution of disapproval, which would block the deal from going forward.
Several members of Congress have already expressed concerns about the agreement because it abandons longstanding conditions for U.S. nuclear cooperation that have bipartisan support. Sen. John Kennedy (R-La.) told the CBS show, “Face the Nation,” July 26 that he was not comfortable with the nuclear cooperation agreement as presented but said, “I don’t think it’s going to happen, anyway.”
All U.S. nuclear cooperation agreements must meet certain nonproliferation criteria enumerated in Section 123 of the Atomic Energy Act, such as the requirement that all nuclear material transferred under an agreement be under IAEA safeguards. Congress, however, has generally supported requiring states to adhere to even more intrusive IAEA monitoring and pushed them to forgo the option to enrich uranium and reprocess plutonium, an approach to nuclear cooperation agreements known as the “gold standard.”
The Trump administration already acknowledged in a November 2025 report to Congress that the nuclear cooperation agreement with Saudi Arabia will not meet the gold standard. The text of the deal confirms that the United States will support Saudi Arabia’s decision to develop uranium enrichment and will not require it to negotiate a more intrusive IAEA safeguards agreement, known as the additional protocol. (See ACT, April 2026.)
The additional protocol, which was finalized in 1997 after illicit nuclear programs in Iraq and North Korea exposed gaps in the NPT-required safeguards, gives the IAEA expanded access to a country’s nuclear sites and additional information about the program. The United States has long supported the universalization of the additional protocol and pushed for its inclusion in nuclear cooperation agreements.
In the November 2025 report to Congress, the Trump administration said that the United States would implement a bilateral safeguards agreement with Saudi Arabia in lieu of the additional protocol. A document outlining the safeguards arrangements submitted to Congress Aug. 24 includes additional details on the information and access that Saudi Arabia will provide the IAEA.
The documents confirm that cooperative U.S.-Saudi nuclear activities could include uranium enrichment, a capability that Riyadh says it wants to develop in order to produce nuclear reactor fuel. (See ACT, March 2023.)
The nuclear cooperation agreement requires a two-year feasibility study that examines the commercial viability of uranium enrichment in Saudi Arabia. If the study supports enrichment, one option could be for the United States to build and operate an enrichment facility in Saudi Arabia to prevent Riyadh from having direct access to U.S. enrichment technology—but it is unclear how Washington would prevent information about enrichment from leaking to Saudi officials or secure the facility in the event of conflict. The United States has never supplied enrichment technology as part of a nuclear cooperation agreement.
According to The Wall Street Journal, if the study does not support enrichment, Saudi Arabia will be free to pursue uranium enrichment domestically or with another foreign partner after 10 years.
Members of Congress have also raised concerns about the implications of the Saudi nuclear agreement on proliferation risks in the Middle East. Sen. John Barrasso (R-Wyo.) warned that the deal could increase proliferation risks in the region and make it more challenging to get Iran to give up uranium enrichment. Sen. Chris Murphy (D-Conn.) said it could “end up pushing Iran closer to a nuclear weapon.”
As executive director for the past quarter-century, Daryl G. Kimball has led the Arms Control Association through setbacks and breakthroughs, gloom and hope in the arms control field.
September 2026
By Thomas Countryman
A couple of years ago, the Arms Control Association celebrated its 50th anniversary as one of the leading nongovernmental organizations dedicated to confronting the risk of planetary suicide by nuclear weapons. We marked last month with an equally significant milestone: Daryl G. Kimball has now led our association as executive director through 25 years of setbacks and breakthroughs, gloom and hope in the arms control field.
Daryl came to this work inspired partly by his family. His mother organized and campaigned tirelessly to demand a cleanup of the consequences of the government’s uranium enrichment program near their home in Ohio. His father was a noted scholar of U.S. nuclear policy.
Daryl first came to Washington in the Scoville Fellowship, the innovative program that—to this day—gives so many scholars and activists their first experience with arms control and nonproliferation issues, and from which the ACA team has drawn several of its top experts.
Daryl started not just in research, but in media and public education efforts that still inspire the association’s efforts to be not just a think tank, but a “think and do” tank. At Physicians for Social Responsibility, his leadership and partnership helped achieve a U.S. Congress resolution endorsing a nuclear testing moratorium in 1992 and engendered administration and congressional support for negotiation of the Comprehensive Test Ban Treaty (CTBT).
Following the dissolution of the Soviet Union, the 1990s were a time when many of us believed that still more significant progress could be made toward a world free of nuclear weapons. In 1997, Daryl became the executive director of the Coalition to Reduce Nuclear Dangers, which brought together 17 of the largest nongovernmental organizations to push for ratification of the CTBT and demand additional reductions in the world’s nuclear arsenals.
Asked by the ACA Board of Directors to become the association’s executive director in 2001, Daryl brought the same focus and drive to an organization already prominent not just for its longevity, but for its credibility and objectivity in analyzing the national security choices facing the United States and the world. The ACA has become a prime forum for senior officials in successive U.S. administrations, and even future presidents, to outline their vision of managing nuclear risk. When we had administrations whose officials were open to dialogue, ACA’s voice was not just heard, but respected, and—yes—influential. In 2011, the association received the MacArthur Award for Creative & Effective Institutions for its ability to address “pressing national and international security challenges with an impact that is disproportionate to its small size.”
All of us who have worked with or for Daryl—and the list of senior officials and creative thinkers who began their careers at ACA is long and impressive—appreciate the asset that best marks a successful leader: genuine appreciation for his coworkers and an active concern for their professional advancement. It matters no less in a small organization than in a lumbering bureaucracy, and it shines through in Daryl’s daily work.
Our challenges today are aggravating: an economic environment that puts the financial squeeze on nongovernmental organizations in national security and other spaces; an ever more crowded public marketplace of ideas, where an even-tempered voice strains to be heard; an erosion of the arms control agreements that have provided a small measure of peaceful breathing to the world; a government disinclined either to listen, or to explain, to the experts whose experience has contributed so much to decades of complex national security issues.
But we do not have the luxury of being discouraged, the time to indulge despair, nor the option to admit defeat. For a quarter-century, we have relied on Daryl Kimball’s judgment, focus, and energy to confront challenges great and small. This would be a good moment for all of our readers to express our thanks to him and honor his significant contributions.
Thomas Countryman, chairman of the board of the Arms Control Association, is a retired U.S. foreign service officer who served as acting undersecretary of state for arms control and international security, among other senior positions.
This analysis was updated on September 1, 2026, to take into account details of the text of the agreement and associated documents after they had been transmitted to Congress.
The danger of nuclear war continues to loom over humanity.
September 2026
By Daryl G. Kimball
The danger of nuclear war continues to loom over humanity. Without new restraints, a more dangerous era lies ahead. The United States, China, and Russia are seeking to modernize and upgrade their arsenals to address perceived nuclear deterrence gaps, including by increasing deployed strategic arsenals and fielding “more usable” tactical nuclear weapons for regional war scenarios.
Into this volatile landscape comes artificial intelligence—not as a future concern, but as a present reality. Private technology companies and governments are caught up in a high-stakes competition to develop advanced frontier AI systems. Leading technologists warned in July that “they could be close to automating AI research,” which will accelerate AI development possibly “beyond our ability to understand or control the resulting systems.”
These two existential challenges—disarming AI and preventing nuclear war—can no longer be understood or governed in isolation.
As nuclear weapons competition and risk grow, the U.S., Chinese, and Russian militaries are racing to integrate AI tools into their conventional military systems and nuclear command-and-control systems to process more information more quickly to provide, as the Pentagon describes it, a “decision advantage” in a crisis or a war.
But the unconstrained use of AI tools may distort the information humans use and the choices they consider when making world-changing decisions regarding nuclear weapons. For example, AI-enabled warfare will compress the time available for meaningful human judgment. Commanders under pressure may be handed algorithm-generated menus of response options that neglect creative, human-centered solutions.
Leaders under stress are also vulnerable to automation bias—the tendency to trust a machine-generated information and analysis over one’s own judgment—even though AI systems are well known for projecting confidence when they are wrong. Another risk is that autonomous weapons with AI-enabled targeting plans could be used to attack nuclear weapons-related infrastructure and prompt a nuclear retaliation. India launched such an attack during its 2025 war with Pakistan.
These and other factors mean that the reckless integration of AI into military and nuclear operations will introduce new uncertainties to the already uncertain and unsustainable practice of nuclear deterrence. The likely introduction of recursive, self-improving, and agentic AI systems in the near future will only increase the risks of catastrophic outcomes.
Unfortunately, the political, diplomatic, and legislative responses to regulate these technologies, and to mitigate the nuclear risks they pose, are failing to keep pace. The best we have, for now, is the growing international consensus that machines must never supplant human judgment in the authorization or execution of nuclear weapons launches.
This “human in the loop” principle, while necessary, is insufficient. There are no agreed definitions, rules of the road, or accountability mechanisms. Nuclear-weapon states have, so far, failed to engage in sustained discussions on how they each plan to put the principle into practice and enforce it.
As a prestigious group of Nobel laureates, technologists, and nuclear experts recommended in an appeal issued from Rome in July, governments, with help from frontier AI developers, researchers, and the international community, need to act now to avoid the reckless integration of AI into nuclear operations.
There are steps that could be pursued to enhance mutual security, starting with a formal commitment by the major nuclear-armed states, through coordinated action or ideally a binding agreement, to retain effective human control over any decision to use nuclear weapons.
These states could also engage in technical talks on measures to insert automated “firebreaks” in advanced command-and-control systems to disallow AI-enabled actions leading to nuclear escalation. By taking concrete steps to move away from postures that call for prompt retaliatory nuclear counterattack and agreeing that the sole purpose of nuclear weapons is to deter nuclear attack, they could also reduce the risk of miscalculation.
Military officials from the nuclear-armed states should be required to report regularly on their use of AI-assisted tools in nuclear operations and the measures that they have implemented to mitigate risks. Independent audits should be conducted to ensure that best practices are being followed.
Progress toward adopting such nuclear risk reduction measures, as well as new and verifiable agreements to cap and reduce nuclear arsenals, is not possible without sustained, serious dialogue among the major nuclear-armed states. Since promising a “new era” of multilateral nuclear arms control after the New Strategic Arms Reduction Treaty expired in February, however, the Trump administration has failed to engage constructively with either China or Russia on nuclear weapons matters.
Although such dialogue may appear unlikely, it is necessary and possible. U.S. President Donald Trump and Chinese President Xi Jinping could set a new and safer course by launching high-level, bilateral talks on nuclear risk reduction and rules of the road for AI use in nuclear weapons systems at their summit this month in Washington.
Immediate and decisive action is required to counter these growing nuclear dangers before it is too late.
In a declaration issued at a conference at the Vatican, the laureates and other experts called for an international treaty prohibiting “reckless integration” of AI in nuclear weapons operations.
September 2026
By Mia Clarke, with additional reporting from Daryl Kimball in Rome
Motivated by the accelerating development of artificial intelligence and multiplying nuclear weapons dangers, Nobel laureates and other experts have called for an international treaty prohibiting “reckless integration” of AI in nuclear weapons operations.
The declaration emerged from a conference July 14-16 at the Vatican. The conference followed the publication of Pope Leo XIV’s path-breaking May 15 encyclical, Magnifica Humanitas, which calls for a global effort to shape and regulate AI in ways that protect human dignity and promote just peace and urges civil society, AI builders, and governments not to leave pivotal decisions in AI policy solely to the technology elites.
In preparatory materials for the conference at the Pontifical Villas in Castel Gandolfo, the Pope’s traditional summer retreat, the organizers asserted that “In the midst of a worsening nuclear arms race, we are embarking on an equally dangerous AI race” and these “great challenges of the twenty-first century can no longer be understood or governed in isolation.”
Called the “global Nobel laureates assembly on artificial intelligence and nuclear war,” the conference brought together a diverse array of more than 200 participants, including Nobel Prize winners in peace, medicine, economics, and physics; AI experts from Anthropic, Open AI, and Google DeepMind; founders and leaders of international peace organizations; nuclear weapons researchers and experts from more than 30 institutions; and religious leaders from several faiths.
The conferees heard expert presentations on key themes including the rapid development of increasingly powerful frontier AI models, the risks posed by the integration of AI in conventional military systems as well as nuclear command and control; the moral and ethical implications of AI adoption on a societal scale; and potential pathways to nuclear conflict.
At the conclusion of the assembly July 16, the participants issued the “Rome Declaration,” which recommends measures that would: ensure the responsible development and governance of AI models so they align with “humanity’s interests, including international law and respect for human rights” and prevent “fully-automated recursive self-improvement in artificial intelligence systems without the means to monitor, and if needed, to halt such systems.”
They called for “an international treaty prohibiting the reckless integration of artificial intelligence into nuclear command, control, and launch systems so that meaningful human control is maintained.” The declaration also urged measures that would “halt renewed arms racing and the expansion of [nuclear] arsenals, resume bilateral and multilateral negotiations on arms control and disarmament” to “diminish the role of nuclear weapons, and reduce the likelihood of first use and catastrophic war.”
The gathering occurred a year after another Nobel Laureates assembly in Chicago on the prevention of nuclear war, which also included a “call on all states to engage in increased cooperative dialogue on the scientific, legal, and military implications of [AI] technologies.”
Reflecting on the importance of these assemblies, 2011 Nobel physics laureate Brian Schmidt said, “It’s a very complicated new space. I think by bringing the groups together, it has clarified the risk.”
But the organizers and participants also recognized that the assembly’s call to action requires urgent follow-up and action. At the July 16 press conference announcing the Rome Declaration, Daniel Holz, a University of Chicago astrophysicist and chair of the Bulletin of the Atomic Scientists’ Science and Security Board cautioned, that “AI is a technology we do not understand, and it’s developing faster than humans can comprehend.”
Shortly after the assembly concluded, more than 1,300 frontier AI developers and corporate leaders issued their own appeal, warning that they “could be close to automating AI research.”
Their July 29 statement, entitled “Pacing the Frontier,” said this achievement could “risk that capability development rapidly accelerates beyond our ability to understand or control the resulting systems.” They end their statement with a vague “request that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.”
Nikita Gryazin, a fellow at the European Leadership Network, wrote in a July 27 commentary about the conference’s outcome document: “The Rome Declaration is a useful starting point, but its practical value will depend on what follows. Governments and international organisations must turn its principles into rules and risk reduction mechanisms, while technology developers must make relevant systems transparent.”
The launch was the second Pacific-facing long-distance Chinese missile test in recent years.
September 2026
By Xiaodon Liang
The Chinese military successfully tested a submarine-launched ballistic missile (SLBM) July 6, delivering a dummy warhead to a target location in the Pacific Ocean, according to state media.
The People’s Liberation Army (PLA) Navy fired the missile approximately 7,000 kilometers from the South China Sea to open waters in the South Pacific, between Kiribati and Tuvalu. Warning notifications issued by Chinese authorities before the test indicate that a second launch from the Bohai Sea, off northern China, was canceled.
The PLA Navy characterized the test as “a routine arrangement … [that] complies with international law and international practice, and is not directed at any specific country or target,” according to a Xinhua report.
The U.S. military also routinely conducts land- and submarine-launched missile tests in the Pacific.
The launch was the second Pacific-facing long-distance Chinese missile test in recent years, following an intercontinental ballistic missile test in late 2024. (See ACT, November 2024.)
Since a pair of Pacific-facing tests in 1980, all of China’s long-distance missile tests have been conducted in the opposite direction, toward targets in western China. These include prior successful tests of land- and sea-based intercontinental missiles, which often occur without notice unless disclosed by other countries.
The missile tested July 6 may have been either a Julang-2 (JL-2), which the Chinese Navy has deployed for a decade, or the newer JL-3, deployed in 2022. The trajectory would have exhausted the range of the JL-2. The JL-3 likely has an extended range of 10,000 kilometers, allowing PLA Navy ballistic missile submarines to range most of the continental United States from littoral waters.
Open-source analysts warn that the JL-2 and JL-3 missiles are difficult to distinguish based on external observation. In July 8 parliamentary testimony, however, a senior Taiwanese intelligence official said the tested missile was a JL-2.
The test was conducted off the coast of Guangdong province from a surfaced vessel in the South China Sea, according to the official, Hsieh Jih-sheng, the deputy chief of intelligence of the general staff of Taiwan’s Ministry of National Defense.
Regional states protested the July 6 test despite receiving prior notification of the event from Beijing.
Japan “strongly urged” China not to conduct the test or endanger Japanese airspace, a government statement said. Although the test trajectory from the South China Sea could not have infringed on Japanese sovereignty, the scrapped prospective test from the Bohai site might have.
“Australia has been clear with China that we regard this as destabilizing to the region,” Foreign Minister Penny Wong told press July 6. She added that the “rapid military buildup by China … [lacks] the transparency and reassurance as to intent that the region expects.”
Matthew Wale, the prime minister of the Solomon Islands, said July 7 that his country does not “want to see any more countries—China, America, anybody—testing in the Pacific Islands region, that’s the bottom line,” according to Australia’s ABC News.
New Zealand Foreign Minister Winston Peters said that “This missile was fired into the South Pacific Nuclear Free Zone established by the Treaty of Rarotonga. China’s action goes against the object and intent of that Treaty,” according to a July 7 report by Radio New Zealand.
He also noted that a pre-launch notification was received only “hours” before the test.
On Aug. 11, the United States and 40 other countries, in a joint statement issued at the Conference on Disarmament in Geneva, called on “all relevant countries” to commit to providing “sufficient advance notice” of at least 24 hours before ballistic missile tests.
The statement, which cites “Recent missile testing activity in the Pacific area,” noted that 145 countries already participate in the Hague Code of Conduct Against Ballistic Missile Proliferation, a political agreement that includes a prelaunch notification obligation.
Although the Hague agreement does not include a 24-hour standard for advance notice of launches, the 1988 U.S.-USSR and 2009 China-Russia launch notification agreements do.
Code adherence have privately discussed and recommended a 24-hour standard, however.
The Aug. 11 statement also called on states to provide in launch notifications “the generic class of the ballistic missile or [space-launch vehicle], the planned launch notification window, the launch area, and the planned direction.” In response to a question regarding criticism of the SLBM test, a Chinese Foreign Ministry spokeswoman said July 7 at a press conference that “there is no need for any country to overinterpret the test launch.”
Syria’s new government also allowed IAEA inspectors access to the site, which was likely part of the former government’s illicit nuclear weapons development program.
September 2026
By Kelsey Davenport
Syria’s new government notified the International Atomic Energy Agency (IAEA) of a newly discovered cache of nuclear materials and allowed agency inspectors access to a site that was likely part of the former government’s illicit nuclear weapons development program.
IAEA Director-General Rafael Mariano Grossi praised Syria’s “courageous decision” to notify the IAEA about the nuclear material in an Aug. 18 joint press conference with Syrian Foreign Minister Asaad al-Shibani in Damascus.
Grossi visited Syria Aug. 17-18 and went to the location where Syria discovered several tons of undeclared nuclear material, which Damascus described as a legacy of the former Assad government. Syria informed the agency that it discovered the material in a July 17 letter. The letter has not been made public but Shibani referenced it during the Aug. 18 press conference.
Grossi did not specify what type of material was found at the site, but Shibani, said that the nuclear “materials do not pose a risk.” Axios reported Aug. 10 that U.S. officials said that the nuclear material is uranium oxide, or yellowcake, a refined uranium ore that can be converted into gas for enrichment. The enriched uranium is then fabricated into fuel for a nuclear reactor.
Under deposed President Bashar al-Assad, Syria conducted undeclared nuclear activities, including the construction of a nuclear reactor at al-Kibar in the Deir el-Zour province. The reactor, which was never declared to the IAEA, was likely intended to produce plutonium for nuclear weapons. Israel destroyed the unfinished reactor in a September 2007 air strike. The U.S. intelligence community later assessed that North Korea assisted Syria in constructing the reactor. (See ACT, April 2018.)
After the Israeli strikes, the Assad government refused to allow the IAEA to inspect the bombed site when the agency first requested access in 2008 and denied conducting undeclared nuclear activities. Syria continued to obstruct the agency’s efforts to investigate its undeclared nuclear activities and later cleared the site. In 2011, the IAEA Board of Governors adopted a resolution finding Syria in noncompliance with its legally required safeguards obligations under the nuclear Nonproliferation Treaty. (See ACT, July/August 2011.)
Before being overthrown in 2024, Assad let an agency team visit the al-Kibar site. The IAEA took samples, which indicated that uranium had been present at the location. In June 2025, Grossi met Syria’s transitional President Ahmad al-Sharaa and announced that Syria’s new government agreed to give the agency unrestricted access. (See ACT, July/August 2025.)
During his Damascus visit, Grossi said the newly discovered nuclear materials will now be accounted for and “put within the international safeguards system.” He added: “In this way, we will be able to work towards the restoration of Syria as a country in full compliance with all its legal obligations in the area of nuclear nonproliferation.”
Shibani said the nuclear materials will “remain in Syria’s custody, while being subject to the agency’s supervision system.” He suggested during the press conference that Syria may seek to use the nuclear materials for a peaceful nuclear program.
Grossi also visited the al-Kibar site while in Syria. He confirmed that the IAEA’s inspection and excavation of the site was already underway and described the results of the cooperative effort as “really, really remarkable.” He said the agency’s work at al-Kibar “validates the long-standing evaluation” that the nuclear reactor at the site “was being built in secrecy.”
An Aug. 18 joint statement from Grossi and Shibani said that Grossi will report the agency’s conclusions to the IAEA Board of Governors “with the aim of a full resolution of the outstanding issues.”
The Syrian government is also cooperating with the Organisation on the Prohibition of Chemical Weapons to dispose of chemical weapons stockpiled by the Assad regime. (See ACT, July/August 2026.)
In two new directives, the administration claims to seek accelerated adoption of advanced AI models by the national security establishment but also incorporates new restraints.
September 2026
By Michael T. Klare
After first giving U.S. technology firms a relatively free hand to speed development of cutting-edge artificial intelligence models, the Trump administration has issued two directives signaling a more cautious, custodial stance.
This new approach was unveiled June 2, with the issuance of an executive order on AI security, and June 6, with a national security presidential memorandum. Both directives claim to seek the accelerated adoption of advanced AI models by the national security establishment, but also incorporate several restraints and guardrails including protections against the release of AI models with undetected flaws and overreliance on technology contractors with their own agendas.
“The national security enterprise shall accelerate AI adoption by identifying mission areas where AI can enhance operational effectiveness and [by] eliminating unnecessary barriers to rapid deployment,” the memorandum asserts, echoing previous administration AI statements.
But, deviating from earlier guidance, it states, “To protect American warfighters … rigorous security and functionality measures, including testing, evaluation, validation, and verification, shall be implemented to assure the appropriate confidentiality, integrity, reliability, availability, and interoperability of AI systems across the national security enterprise.”
The directives are the latest shift as the Trump administration struggles to devise a clear policy on government oversight and use of the technology as the computing power of advanced “frontier” AI models has increased.
The June 2 executive order authorizes the National Security Agency to determine when an advanced AI system reaches the threshold of a “covered frontier model” and to establish a mechanism whereby developers of those systems voluntarily submit them to the federal government for independent review prior to public release, enabling timely repair of any flaws detected in the process.
The June 6 memorandum bans the use of advanced AI to “conduct unauthorized or unlawful surveillance activities” on U.S. citizens and empowers the government to terminate contracts with technology suppliers “that have repeatedly demonstrated a pattern of conduct that is inconsistent with” administration policy.
The administration’s earlier directives eschewed such restrictions. In his first initiative, Executive Order 14179 of January 23, 2025, President Donald Trump enjoined senior administration officials to terminate all measures adopted by the Biden administration to regulate advanced AI models.
The Trump administration’s original approach to AI development was further articulated in “Winning the Race: America’s AI Action Plan,” released in July 2025. The White House then viewed AI innovation as a race—primarily with China—and believed that a hands-off policy on commercial AI advancement would best ensure a U.S. victory. “The United States is in a race to achieve global dominance in artificial intelligence,” the plan states. “The United States needs to innovate faster and more comprehensively than our competitors … and dismantle unnecessary regulatory barriers that hinder the private sector in doing so.”
But recent developments caused the administration to rethink its stance on dismantling regulatory barriers to AI development, including evidence of rogue behavior by advanced AI models—instances when they circumvented guardrails imposed by human programmers to prevent unauthorized functions—and disputes over the roles private firms play in setting AI functions.
“The political winds have shifted in our direction on this, mostly because the capabilities are increasing, and there are concerns about advanced AI without safeguards,” Brendan Steinhauser, the chief executive of the nongovernmental organization Alliance for Secure AI Action, told The New York Times in February.
Among the factors affecting the Trump administration’s thinking is its dispute with Anthropic over the AI firm’s attempt to impose limits on how the U.S. Defense Department might employ its Claude program in military operations. During contract negotiations in February, Anthropic insisted that the department eschew using Claude to govern autonomous weapons systems or to conduct mass surveillance of U.S. citizens. The move incensed Defense Secretary Pete Hegseth, who said that only the Pentagon could set the rules for its use of AI products and declared Anthropic a “supply chain risk,” essentially barring it from federal contracts.
Although many federal agencies continue to employ Claude in one capacity or another and the Anthropic-Pentagon dispute, now in federal court, may yet be resolved amicably, U.S. officials seem determined to prevent future instances of corporate overreach in this area, as shown by the June 6 memorandum, with its mandate for the termination of any AI programs with unauthorized corporate provisions.
In a second challenge to the administration’s earlier approach, another Anthropic tool, Claude Mythos, was revealed by the company to be highly proficient at detecting flaws in existing computer code, potentially enabling malign operators to use the system to hack into sensitive computer systems and inflict serious harm.
At first, Anthropic and the Trump administration sought to restrict access to Mythos, but it soon became obvious to AI experts that similar capabilities were, or would soon be available, from other tech companies, including some in China. Beijing “is already racing to build and acquire these AI capabilities and may be closer to developing its own Mythos than many U.S. policymakers hope,” Michael Sulmeyer, a former assistant secretary of defense for cyber policy, wrote in Foreign Affairs.
Defense against advanced AI-enabled cyberattacks thus has become a national security priority, as reflected in Trump’s June 2 executive order, which calls for an “AI cybersecurity clearinghouse” intended to facilitate government-industry collaboration in detecting and repairing key cyber vulnerabilities.
German Chancellor Friedrich Merz said the exercise would be held by the end of the year.
September 2026
By Libby Flatoff
German Chancellor Friedrich Merz announced that the German military will be participating in a French nuclear exercise for the first time as a part of France’s new deterrence strategy.
“We will have German conventional forces participate in a nuclear exercise conducted by the French armed forces before the end of this year,” Merz said at a July 17 news conference with French President Emmanuel Macron, according to the Associated Press.
The announcement came after Germany and France formed a nuclear steering group March 2, “to coordinate strategic doctrine and exercises.” (See ACT, March 2026.)
Politico reported July 16 that the drills would be part of France’s flagship nuclear exercise, “Poker,” and take place in September. The Federation of American Scientists reported in its 2025 overview of French nuclear weapons that France’s Strategic Air Forces Command conducts about 70 exercises per year, four of which are part of Poker.
The exercise involves “a majority of France’s nuclear-capable Rafale aircraft, which during flights carry unarmed air-sol moyenne portée-amélioré (ASMPA) air-launched cruise missiles,” according to the federation’s profile. The drill carries out a simulated nuclear aid raid and includes refueling operations by A330 MRTT tankers and “penetration of enemy airspace and firing of ASMPA missiles by Rafale fighters.”
On July 16, France and Germany took the first operational step and conducted a joint conventional air activity “in the context of the French concept of forward deterrence,” according to a press release from the French armed forces ministry. No nuclear forces were used, the release stated.
At the July 17 news conference, Merz emphasized that this new drill “complements our commitment to NATO’s nuclear sharing and deterrence arrangements, which we continue to uphold.” Macron noted that he viewed Germany as playing a “vanguard role” in European deterrence.
In April, Germany produced its first defense strategy that noted its goal is to become “the strongest conventional
army in Europe.” The strategy names Russia as Europe’s biggest threat and also asserts that “Germany must … become an even stronger military Ally to the United States.”
As of January 2026, Stockholm International Peace Research Institute estimated that a number of NATO member states, including Germany, together host 100 to 120 U.S. B61-12 nuclear bombs as a part of the alliance’s nuclear sharing agreement.
The Franco-German exercise will occur after the United States announced a new six-month force posture review in Europe as it pushes Europe to be primarily responsible for its own defense (See ACT, July/August 2026.) France also conducted two other nuclear steering group meetings, one with Sweden April 23 and the other with Poland July 10.
Twelve years after reaching a nuclear cooperation agreement, the two countries finalized the technical arrangements to allow exporting to begin.
September 2026
By Kelsey Davenport
Twelve years after Australia and India reached a nuclear cooperation agreement, the two countries finalized the necessary technical arrangements to allow Australia to begin exporting uranium to India.
Australian Prime Minister Anthony Albanese and Indian Prime Minister Narendra Modi announced the deal during the third annual India-Australia summit in Melbourne July 9.
The agreement, which took two years to negotiate, provides a framework for implementing the 2014 nuclear cooperation agreement. According to the Australian government, it addresses long-standing conditions that any exported uranium must be used solely for India’s civil nuclear program and not for nuclear weapons by establishing tracking and accountancy procedures. Neither Australia nor India has provided details about how the accountancy arrangement will function.
In a joint press conference July 9, Albanese emphasized that Australian uranium would be used only for “peaceful purposes” in India and “help increase the share of non-fossil fuel power capacity.” Modi said the deal “will pave the way for uranium supplies from Australia to India and give our clean energy objectives fresh momentum.”
India operates 24 nuclear reactors and is building 10 more, but it relies on imports to fuel its expanding nuclear energy program. Modi initially sought a nuclear cooperation agreement with Australia to secure necessary uranium for fuel purposes. Australia is the fourth-largest producer of uranium and it leads the world in uranium reserves, at about 28 percent.
When the two countries reached a nuclear cooperation agreement in 2014, Modi described the deal as a “historic milestone” and said it would provide his country with “long-term reliable supplies of uranium.” (See ACT, October 2014.)
But after the nuclear cooperation agreement was finalized, Australia’s Joint Standing Committee on Treaties recommended that the country delay any uranium export contracts until certain conditions were met, including on safeguards and inspections, to ensure that Australian uranium did not end up in India’s nuclear weapons program.
As a member of the Treaty of Rarotonga, which established a nuclear-weapon-free zone in the South Pacific, Australia cannot sell nuclear materials to any country that does not have a safeguards agreement with the International Atomic Energy Agency (IAEA). The treaty specifically states that the recipient state must have a safeguards agreement as required by the nuclear Nonproliferation Treaty (NPT). India, which developed nuclear weapons outside of the NPT, does have an IAEA safeguards agreement, but it only applies to certain civil facilities.
When the Australian Parliament voted to lift its ban on selling uranium to India in 2011, critics of the deal argued that the facility-specific safeguards are not the same as the comprehensive safeguards agreement required by the NPT. India signed a facility-specific additional protocol with the IAEA in 2014, which helped facilitate its nuclear cooperation agreement with Australia. The additional protocol is designed to provide the IAEA with expanded access to information and sites, but India’s additional protocol omits some of the key transparency provisions included in the IAEA’s model agreement. (See ACT, March 2009.)
