AI is rewriting the rules of how governments operate, regulate, and wage war. The current coverage falls into two buckets: too technical for the people making decisions, or too shallow for the people implementing them. If you're a program manager at a civilian agency, a defense contractor, a consultant helping agencies modernize, or an operator trying to understand how AI policy affects your business, you need something different.

That's what this newsletter delivers. Twice a week, I'll cut through the noise to give you the context, the analysis, and the "so what" behind AI in government.

I've spent my career at the intersection of business strategy, technology, and value outcomes, leading AI strategy and go-to-market efforts for enterprise SaaS companies. Before that, I spent a decade in management consulting guiding organizations through large-scale transformations. Since 2004, I've had a front row seat to every major technology shift that has reshaped how organizations operate: the rise of enterprise SaaS, the migration from on-premise to cloud, the mobile-first wave that changed how workforces and customers interact, and the push into big data and analytics that promised to make every organization "data-driven." Each of those shifts followed a similar pattern. Early hype, messy adoption, real winners, and a lot of expensive lessons. AI is following that same arc, but faster, with higher stakes, and with far less clarity on the rules.

Here's how The AI Mandate works.

Every Tuesday, I send The AI Mandate Scan: a curated rundown of the most important AI-in-government news, policy moves, and use cases worth watching.

Every Thursday, I publish The Intelligence Report, where I go deeper on one or two topics that matter.

This is the first Intelligence Report. Today’s Report includes:

Fair warning: this inaugural edition runs longer than a typical Thursday will in the future. I believe the stories covered here warrant the extra space.

Let's get into it.

Operation Epic Fury launched on February 28. Within the first 24 hours, U.S. Central Command struck over 1,000 targets across Iran, a pace that military leaders have attributed directly to AI-assisted systems. Admiral Brad Cooper, head of CENTCOM, confirmed the use of "a variety of advanced AI tools", stating that these systems help warfighters "sift through vast amounts of data in seconds" so commanders can "make smarter decisions faster than the enemy can react."

This is the largest U.S. military operation since the 2003 Iraq invasion and AI is embedded in the operational core.

The centerpiece of the U.S. military's AI integration is the Maven Smart System, developed by Palantir in partnership with the Pentagon. Maven fuses satellite imagery, drone feeds, radar data, and signals intelligence into a unified platform that classifies targets, recommends weapons, and generates strike packages in near real time. What used to take hours or days now takes minutes.

Anthropic's Claude large language model is embedded within the system to summarize intelligence, analyze data, and simulate scenarios. Yes, the same Claude that the Pentagon designated a "supply chain risk". The irony is hard to miss: the military is relying on technology from a company it’s trying to blacklist.

The operational tempo tells the story. According to defense sources cited by IBTimes, analysts are processing roughly 1,000 targets daily with about 10% of the human resources previously required. A 2024 Georgetown University investigation found the U.S. Army's 18th Airborne Corps had used AI to reduce an intelligence processing team from 2,000 people to just 20. The "kill chain," the time between identifying a target and striking it, has been compressed from weeks (in World War II) to minutes.

Military officials insist that humans retain final authority over lethal decisions. Cooper stated plainly: "Humans will always make final decisions on what to shoot and when to shoot." But as Lauren Kahn of Georgetown's Center for Security and Emerging Technology told NPR, the more dangerous challenge is the "blurring of where the AI starts and where the AI stops." When a system generates a target recommendation in seconds and a human approves it in seconds, the distinction between "AI-assisted" and "AI-driven" starts to thin.

AI's role extends beyond targeting. U.S. Cyber Command used AI-enabled operations in the war's early hours to disrupt Iranian command, control, and communications networks. Iran retaliated by striking AWS data centers in the UAE and Bahrain, aiming to impair the commercial cloud infrastructure that supports U.S. military AI systems. That's a significant escalation: cloud infrastructure is now a military target.

On the information front, Brookings has documented an unprecedented volume of AI-generated disinformation flooding social media since the conflict began, including fabricated footage of explosions in Tel Aviv, fake satellite imagery, and synthetic video recycled from other conflicts and even video games (which were shared on official White House social media accounts). Chinese firms have been marketing AI-derived intelligence detailing U.S. base equipment, carrier group movements, and strike assembly patterns.

Meanwhile, MIT Technology Review reports that dozens of AI-powered intelligence dashboards have sprung up, many "vibe-coded" in days using AI tools, combining open-source data with prediction markets where people bet on war outcomes. One was built by employees at Andreessen Horowitz and caught the attention of a Palantir founder. The line between intelligence analysis, entertainment, and speculation is collapsing.

The Iran conflict is functioning as a live proving ground for military AI, and the results are landing on both sides of the ledger. The operational gains have been significant: Iranian drone attacks have dropped 83% and ballistic missile launches have fallen 90% since the campaign began. Project Maven, launched in 2017, has evolved from a pilot into a program of record. NATO has acquired its own version. Deputy Secretary of Defense Steve Feinberg recently signed a memo formalizing AI's role in military decision-making and pushing Maven adoption across all U.S. military branches by September. Pentagon AI adoption has crossed from aspirational to operational.

But this is also the first large-scale deployment of AI targeting technology in active combat and the early returns raise real questions about deploying unproven systems at this scale.

Maven's reported accuracy hovers around 60%, compared with 84% for human analysts. In adverse conditions like bad weather or poor visibility, that accuracy drops below 30%, according to Pentagon data. In 2021, an experimental Air Force targeting AI scored roughly 25% accuracy in real conditions despite rating its own confidence at 90%. These are baseline performance metrics for a system now processing thousands of targets in an active war.

The Minab school strike crystallizes the tension. On February 28, the first day of the war, a Tomahawk cruise missile struck Shajareh Tayyebeh elementary school in southern Iran, killing at least 168 people, more than 100 of them children. The school sat near a former IRGC compound that had been closed for roughly 15 years. It had its own website and an active social media presence. A Reuters investigation found that outdated intelligence, potentially a decade old, likely contributed to the targeting failure. Former military officials speaking to Semafor pointed to stale human-curated data as the cause, not AI malfunction. More than 120 Democratic members of Congress have written to Defense Secretary Pete Hegseth demanding answers about whether Maven was used to identify the school as a target and whether a human verified the accuracy.

The deeper concern is what experts call "automation bias": if operators trust machine recommendations faster than they can independently verify them, the human in the loop becomes a rubber stamp. As the Bulletin of the Atomic Scientists argued, the military hasn't conducted empirical studies on whether AI is actually reducing or increasing collateral damage. And the institutional safeguards are thinning. Hegseth slashed the Civilian Protection Center of Excellence workforce by approximately 90%, the Pentagon unit responsible for developing policies and tools to minimize civilian harm during military operations, and cut CENTCOM's civilian casualty assessment team from 10 to one.

The Pentagon's own AI strategy document states that "the risks of not moving fast enough outweigh the risks of imperfect alignment." Whether that calculus holds up under scrutiny is one of the defining questions, and one I’ll continue to track here.

Three weeks ago, the White House released its National Policy Framework for Artificial Intelligence, a four-page set of legislative recommendations addressed to Congress.

The framework's most consequential position remains its call for federal preemption of state AI laws that impose "inconsistent or undue burdens." David Sacks, the White House AI czar who co-authored the framework, has been making the rounds on Capitol Hill and told Bloomberg he sees a path to bipartisan AI legislation within months. Sacks framed the framework around what he calls the "5 Cs": child safety, communities, creators, censorship, and competitiveness.

This has led to a massive advocacy blitz around the framework, and the battle lines are forming fast. A new pro-AI group called Innovation Council Action, backed by Sacks and former White House deputy chief of staff Taylor Budowich, plans to spend $100 million to oppose AI regulation. On the other side, a new nonprofit investing at least $10 million in online child protections has already held 15 Capitol Hill meetings pushing for stronger safeguards than the framework proposes.

On the same day the framework dropped, my former representative, Rep. Don Beyer (D-Va.) and four Democratic colleagues introduced legislation to repeal the White House's AI preemption executive order, arguing states need the ability to enact "commonsense safeguards" while Congress debates. On the Republican side, Sen. Marsha Blackburn (R-Tenn.) released a 291-page TRUMP AMERICA AI Act that takes a more prescriptive approach than the four-page White House framework, seeking to codify elements of Trump's AI executive orders while constraining states' regulatory authority. Senate Commerce Chairman Ted Cruz (R-Texas) hasn't committed to the Blackburn approach, and past disagreements over preemption strategy suggest multiple legislative pathways remain in play.

Meanwhile, more than 50 Republicans wrote to Trump expressing concern that federal attempts to halt state AI legislation go too far. Bipartisan resistance to broad preemption is real.

While this fight takes place in D.C., states aren't waiting. Colorado's AI Act takes effect June 30, 2026. California has multiple AI transparency laws with 2026 effective dates. Texas's Responsible AI Governance Act is already enforceable. The Commerce Department still hasn't released its required evaluation of "onerous" state AI laws, which was due March 11. That delay introduces uncertainty about the administration's near-term posture on enforcement.

Alvarez & Marsal's analysis frames the practical reality well: the framework remains a policy recommendation without legal force. The current compliance landscape hasn't changed. Companies remain subject to an uneven mix of state requirements, federal enforcement through existing authorities, and political uncertainty over whether Congress will adopt any of this.

The framework is also notably silent on military and intelligence uses of AI, which is striking given that the Iran conflict has put AI-enabled warfare on the front page. No procurement standards. No testing mandates. No transparency requirements for how agencies evaluate the safety of AI systems deployed in sensitive contexts. As Operation Epic Fury demonstrates, this is the space where the hardest questions about AI governance live.

Despite Sacks's optimism, Davis Wright Tremaine's assessment is blunt: given the number of issues in play, the lack of consensus on how to resolve the conflicting interests, and the fact that 2026 is an election year, the likelihood of near-term Congressional adoption "is remote." Nixon Peabody agrees: near-term federal AI laws are more likely to be targeted (child safety, online harms) than comprehensive. Broad preemption remains politically complex.

A quick scan of how governments around the world are building, deploying, and betting on AI.

🇸🇦 Saudi Arabia designated 2026 as the "Year of AI" and is backing it with real money. HUMAIN, the PIF-owned AI company, announced a joint venture with AMD and Cisco to deploy up to 1 GW of AI infrastructure by 2030. The kingdom's AI sector attracted $9.1 billion in funding through 70 deals in 2025, and a 480-megawatt data center broke ground in January.

🇨🇳 Chinese firms are marketing AI-derived military intelligence (as reported earlier) pulled from commercial satellite imagery and open-source data, detailing U.S. base equipment and carrier group movements during the Iran conflict. Companies like MizarVision hold Chinese military standard certificates and published imagery of U.S. air bases days before they were struck. Sovereign AI isn't just about economic competitiveness. It's about who controls the intelligence layer.

🇮🇳 India unveiled three sovereign AI models at the India AI Impact Summit in February, including Sarvam AI's 105-billion-parameter model trained entirely on domestic infrastructure. The government's IndiaAI Mission is backing 12 organizations with over $1 billion to build indigenous foundation models, 38,000 GPUs of shared compute capacity, and sovereign data platforms across healthcare, agriculture, governance, and education.

One Thing Worth Your Time. DHS publishes a complete inventory of every AI use case deployed across the department, updated January 2026. It now lists over 200 active applications, up 37% since mid-2025. The list includes everything from:

Each use case is categorized by risk level, deployment status, and the specific DHS component using it. The full dataset is downloadable as a spreadsheet. If you want to understand how AI is actually being used inside federal agencies, start here.

That's it for the first Intelligence Report. If this landed for you, share it with a colleague who works at the intersection of AI and government. If you have feedback, questions, or tips, reply to this email. I read everything.

See you Tuesday for The AI Mandate Scan.

— Hamaad