10-minute read | 2,000 words
What to know this week
Anthropic rejects Pentagon’s terms for AI use.
After days of debates, Anthropic announced that it has rejected the Pentagon’s ultimatum for full access to its artificial intelligence (AI) tool, Claude.
Conflict in Iran spills into the cyberverse.
With the Iran conflict continuing to expand, cyber operations are growing in both severity and scope.
This week's full stories
Anthropic has rejected the Pentagon’s ultimatum.
THE NEWS
On Thursday, Anthropic’s CEO, Dario Amodei, announced that it would not concede to the Pentagon’s ultimatum regarding access to its AI tool, Claude. This announcement comes after Secretary of Defense Pete Hegseth imposed a deadline for the company to comply with the Department’s policy or face being cut from future military contracting work. Amodei added that current AI systems are not reliable enough to be used to power weaponry, and current laws need to be updated to account for AI snooping tools.
This clash has been brewing for several weeks after the two sides cited different accounts regarding a discussion about the military’s use of lethal autonomous weapons. According to Pentagon officials, the two sides engaged in a hypothetical conversation about defending the United States (US) from an intercontinental ballistic missile attack using Anthropic’s Claude AI system. Pentagon officials characterized Amodei’s response as suggesting that Anthropic would work with the Department on the matter if needed rather than proactively.
However, Anthropic has pushed back on these claims, stating that the CEO never gave that response. Additionally, Anthropic also stated that the company has already agreed to allow the military to use Claude for missile defense. Instead, Anthropic has stated it is concerned about the military using its AI systems for autonomous weapon systems and the mass surveillance of US citizens.
Sean Parnell, a chief spokesman for the Pentagon, released a statement, writing:
“This is a simple, common-sense request that will prevent Anthropic from jeopardizing critical military operations and potentially putting our warfighters at risk.”
Dario Amodei also released a statement emphasizing:
“In a narrow set of cases, we believe AI can undermine, rather than defend, democratic values. Two such use cases have never been included in our contracts with the Department of War, and we believe they should not be included now.”
Following Anthropic’s rebuttal of the Pentagon’s demands, President Trump directed all federal agencies to “immediately cease” using Anthropic.
Additionally, shortly after ending this deal Pentagon formalized a new deal with OpenAI. More on this story below.
THE KNOWLEDGE
As this matter has become increasingly heated in the past several days, it brings significant attention to how AI systems are being utilized for military purposes.
A key portion of the Trump administration’s AI action plan involves supporting the development of AI systems with “applications to defense and national security.” For example, the US Air Force has been actively testing AI-piloted aircraft, finding that these systems were oftentimes beating elite pilots by improving on turns and maneuvering capabilities.
Behind these efforts to adopt AI for military uses is the US’s competition to outpace similar efforts being undertaken within China. In a Brookings Institution report, Senior Fellow Elsa Kania assessed how the People’s Liberation Army (PLA) is approaching this AI and machine learning arms race. Some of the key projects currently being explored include:
- Developing both autonomous submarines and unmanned surface vehicles capable of operating with some autonomy.
- The PLA’s Rocket Force is augmenting its missile systems to become more “intelligentized.”
- Investing in swarming technologies for drones.
- The PLA’s Strategic Support Force potentially utilizes AI to support its missions in space, cyber, electronic, and psychological warfare.
While nations are still far from being able to field fully autonomous weaponry, this is a goal that many national security agencies are working to achieve. As these efforts continue to develop and advance, this case, as well as future ones, may be instrumental in setting the stage for how the US government will approach utilizing privatized AI services within the military.
THE IMPACT
This dispute represents far more than a contractual disagreement between the Department of Defense and Anthropic. Rather, it is an early test of who will define the rules governing the integration of frontier AI systems into the US’s military operations.
If the Pentagon ultimately relaxes its demands on Anthropic, it could signal that frontier AI developers will be able to retain meaningful authority in how their systems are being deployed by the military. This outcome would set a significant precedent for other AI developers to impose similar restrictions, allowing private contracting decisions to meaningfully shape US autonomous weapons policies in the absence of formal legislation.
However, if the Department of Defense does permanently cut off Anthropic, this could accelerate efforts to incorporate AI systems into defense systems, potentially with fewer internal guardrails to prevent AI misuse.
While today’s AI systems remain unreliable for military environments, these capabilities are advancing rapidly. The outcome of this dispute over the next few months could shape not only how the federal government makes procurement decisions but also significantly affect the power balance between the military and emerging technology companies.
Iranian conflict expands into cyberspace.
THE NEWS
On Monday, reports emerged that both sides of the Iranian conflict have begun to utilize cyberspace to launch a range of attacks on each other. In these reports, both the US and the Israelis launched several cyber attacks throughout Iran. These attacks were able to cause widespread disruptions, including targeting critical communication infrastructure facilities, news and propaganda sites, digital government services, and local applications. Additionally, these cyberattacks also targeted the Islamic Revolutionary Guard Corps (IRGC) command and control systems to limit the nation’s ability to counterattack.
When launching these attacks, the US and Israeli cyber forces reportedly used Distributed Denial of Service attacks alongside “deep intrusions” into energy and infrastructure systems.
From Iran and pro-Iran threat actors, reports have emerged that their operations have scaled up since the conflict began. Iranian threat actors have been targeting air defense systems, fuel infrastructure systems, and other industrial control systems across the Gulf region. These attacks have focused on launching DDoS attacks alongside engaging in data-wiping operations for military logistics providers.
Adam Meyers, the head of counter-adversary operations at CrowdStrike, released a statement emphasizing:
“These behaviours often precede more aggressive operations. In past conflicts, Tehran’s cyber actors have aligned their activity with broader strategic objectives that increase pressure and visibility at targets, including energy, critical infrastructure, finance, telecommunications, and healthcare.”
US cybersecurity firm, SentinelOne, echoed similar sentiments, stating:
“We assess with high confidence that organizations in Israel, the [US], and allied nations are likely to face direct or indirect targeting - particularly within government, critical infrastructure, defense, financial services, academic, and media sectors.”
THE KNOWLEDGE
In the last conflict with Iran in 2025, when both the US and Israel targeted the nation’s underground nuclear sites, US officials warned that Iran would launch retaliatory cyberattacks. At the time, the Department of Homeland Security (DHS) issued a National Terrorism Advisory highlighting the increased chance of “low-level cyber attacks” against US networks. However, the DHS was primarily concerned with attacks being launched by “pro-Iranian hacktivists” rather than coordinated, state-sponsored efforts.
For over a decade, Iran has treated its cyber operations as a tool for power projection. These efforts notably scaled up following the 2010 Stuxnet incident, where the US sabotaged Iranian nuclear centrifuges. From there, state-sponsored cyber units have begun to conduct disruptive and destructive operations targeting financial institutions, energy infrastructure, and industrial systems around the globe.
For example, in October 2023, an Iranian-affiliated group, known as CyberAv3ngers, targeted US water and wastewater systems. The attack compromised at least seventy-five devices, all made by the Israeli company, Unitronics. While initial reports suggested that hacktivists were behind these attacks, cybersecurity organizations pushed back, stating that the funding and tooling needed to execute the operation exceeded normal hacktivist capabilities and were rather indicative of state-sponsored actors.
These cyber operations allow Iran to impose cost, signal resolve, and pressure adversaries while still being able to maintain plausible deniability. In doing so, they can compensate for limitations in Iran’s traditional military capabilities. Given these capabilities, it is not surprising that Iranian cyber operations have become more prominent as hostilities have intensified.
THE IMPACT
This conflict's expansion into cyberspace marks more than a simple retaliation effort; it demonstrates how both sides are committing a significantly larger number of resources.
For Iran in particular, the nation’s cyber capabilities act as a key equalizer. While Tehran cannot conventionally match the US or Israeli militaries, its cyber units provide a significant means to project power regionally and globally. This power will likely continue to be used to pressure industries throughout the Gulf region and to impact targets across the globe.
Depending on how sophisticated these attacks are, the US’s critical infrastructure defenses and resiliency systems will likely be tested over the coming weeks, depending on how this conflict progresses. Further, given Iran’s attacks against other Gulf nations, it is possible that potential cyber attacks could target other nations outside of the US and Israel.
This Week's Caveat Podcast: The Pentagon’s AI ultimatum.
Ben Yelin sits down with N2K’s Lead Analyst, Ethan Cook, to discuss the fallout between the Pentagon and Anthropic. Throughout the episode, the two discuss what caused this relationship to fall apart, how the fallout played out, and how this incident will impact the larger AI ecosystem.
OTHER NOTEWORTHY STORIES
Judge blocks Virginia law restricting social media.
What: A federal judge blocked a Virginia state law, which looked to impose age verification and usage requirements for minors engaging with social media sites.
Why: On Friday, US District Judge Patricia Tolliver Giles blocked Virginia’s social media ban law. With this ruling, the judge stated that NetChoice, a technology trade group, was likely able to establish that the state law was unconstitutional as it would infringe upon free speech rights.
More specifically, Judge Giles wrote:
“The court recognizes the Commonwealth’s compelling interest in protecting its youth from the harms associated with the addictive aspects of social media. However, it cannot infringe on First Amendment rights, including those of the same youth it aims to protect.”
Virginia has argued that this law was designed to protect children from social media’s addictive nature and to begin addressing the youth mental health crisis.
Paul Taske, the co-director of NetChoice Litigation Center, commented on the decision, stating:
“This ruling reaffirms that the government cannot ration access to lawful speech, even if it has noble intentions. Fundamentally, parents must stay in the driver’s seat when it comes to decisions about their families.”
FEB 27, 2026 | Source: Reuters
Australia signals potential intent to target app stores.
What: Australia’s internet regulator signaled its intent to crack down on search engines and app stores.
Why: On Monday, Australia’s regulators stated that they are considering imposing greater regulations on both search engines and app stores to better block AI services that do not verify user ages before engaging with them. This announcement comes after Australia recently imposed a nationwide ban on social media platforms from engaging with minors under the age of sixteen.
A spokesperson for the commissioner stated:
“eSafety will use the full range of our powers where there is non-compliance, [including] action in respect of gatekeeper services such as search engines and app stores that provide key points of access to particular services.”
MAR 1, 2026 | Source: Reuters
Pentagon and OpenAI reach new deal.
What: The Pentagon and OpenAI reached a deal to allow the defense organization to use the company’s AI models.
Why: Over the weekend, the Pentagon and OpenAI established a new deal. In this deal, OpenAI stated that the Department of Defense agreed to prohibitions on using AI for domestic surveillance and autonomous weapons.
This deal follows a significant incident between the Pentagon and Anthropic regarding how the defense agency could and could not use the company’s AI systems due to concerns related to these prohibitions.
Sam Altman released a statement, writing:
“Two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems.”
FEB 28, 2026 | Source: The Hill