Go back

AI Daily Digest - 2026-03-15

AI Daily Digest - 2026-03-15 slide 1

Table of contents

Open Table of contents

US Army Awards Anduril Unprecedented $20 Billion AI Defense Contract

Category: Defense & Enterprise AI Source: TechCrunch

The U.S. Army announced late Friday a massive 10-year contract with defense tech startup Anduril that could reach $20 billion in value, marking one of the largest technology-focused military contracts in recent years. The deal consolidates what had been more than 120 separate procurement actions into a single enterprise agreement covering Anduril’s hardware, software, infrastructure, and services.

Founded by Palmer Luckey—who previously sold VR startup Oculus to Facebook before his controversial departure—Anduril has positioned itself at the forefront of autonomous military technology. The company brought in approximately $2 billion in revenue last year and is reportedly in talks to raise funding at a $60 billion valuation. According to the Army’s announcement, the contract includes a five-year base period with options to extend for an additional five years, focusing on rapid software deployment capabilities for modern battlefield applications.

“The modern battlefield is increasingly defined by software,” said Gabe Chiulli, CTO at the Department of Defense’s Office of the Chief Information Officer. “To maintain our advantage, we must be able to acquire and deploy software capabilities with speed and efficiency.” The deal comes as Anduril’s vision for remaking the U.S. military with autonomous fighter jets, drones, and submarines has found favor with the current administration, though it also highlights growing tensions in the AI defense sector following recent disputes between the Pentagon and companies like Anthropic and OpenAI over military contracts.

Colleges Revive Blue-Book Exams as AI Cheating Concerns Reshape Education

Category: Education & Policy Source: Axios

Universities across America are bringing back blue-book exams—handwritten, in-person tests—in response to widespread concerns about AI-generated writing, but educators are deeply divided over whether this represents sound pedagogy or a regressive retreat. The move comes as ChatGPT and similar tools have fundamentally altered academic integrity discussions since OpenAI’s release in late 2022.

Proponents argue that blue books help prevent copy-and-paste cheating, but critics say the solution is both impractical and unfair. Professor Dan Melzer at UC Davis notes that educators won’t be able to “completely outsmart ChatGPT” as students find workarounds, while Professor Steven Krause at Eastern Michigan University argues the pervasive myth of widespread cheating is overstated. “In my experience, most cheaters are failing and are desperate for solutions,” Krause told Axios, adding that he reads about 1,500 pages of student writing each semester and can usually detect AI-generated content.

The approach faces significant practical challenges: more than half of students now take at least one online course, making in-person exams impractical; multilingual writers and students with disabilities face serious disadvantages in timed handwritten scenarios; and the method isn’t scalable for large classes. Perhaps most critically, educators point out that writing is meant to be a revision process—forcing students to produce timed, single-draft responses evaluates rushed thoughts rather than actual skills. “Out of step and out of date with the way literacy is currently practiced in the world, outside of the ivory tower,” Melzer concluded, noting that many employers actually want graduates comfortable using AI tools.

Meta Delays Next AI Model After Internal Tests Show Performance Gap

Category: Model Development Source: New York Post

Meta is delaying the release of its next artificial intelligence model—codenamed “Avocado”—by at least two months after internal testing revealed it falling short of rival models from Google, OpenAI, and Anthropic, according to a New York Times report. The setback represents a significant challenge for Mark Zuckerberg’s aggressive AI strategy, which has seen the company spend billions on talent acquisition and commit $600 billion to building data centers.

Meta’s AI division is now reportedly considering temporarily licensing Google’s Gemini to power its products while working to improve Avocado’s performance. Internal tests showed the model underperforming in critical areas including reasoning, coding, and writing—failing to measure up even to Google’s Gemini 2.5 from March, let alone the newer Gemini 3.0 from November. The delayed model is now expected to launch around May instead of this month.

The timing is particularly challenging given Meta’s massive financial commitments to AI development. The company projected in January it would spend as much as $135 billion this year alone—nearly double the $72 billion spent last year. Zuckerberg said in July that Meta’s new AI models would “push the frontier in the next year or so,” and emphasized in January that “I expect our first models will be good, but more importantly will show the rapid trajectory we’re on.” The delays raise questions about whether Meta can maintain competitive positioning against rivals who continue to advance their models at a rapid pace.

AI Surveillance Expansion and Water Crisis Spotlight Regulatory Gaps

Category: Policy & Infrastructure Source: The Guardian

Multiple developments this week highlighted growing concerns around AI’s societal impacts, from government surveillance capabilities to environmental resource consumption. Hacked data from the Department of Homeland Security’s technology incubator revealed significant spending on partnerships aimed at expanding surveillance capabilities using artificial intelligence, raising fresh privacy concerns about the scope of government monitoring programs.

Separately, Washington state’s defeat of House Bill 2515 exposed the tech industry’s lobbying power and the environmental costs of AI infrastructure. The state hosts approximately 126 AI data centers that evaporate millions of gallons of freshwater daily for cooling while consuming enough electricity to risk becoming the single largest power consumer in the Pacific Northwest. The bill would have required facilities to use clean energy and reduce power during high demand periods to protect ratepayers and prevent blackouts.

“Water used for AI computing is mostly ‘blue water’—fresh water from rivers and lakes, which humans could use directly,” explained Shaolei Ren, an electrical engineering professor at UC Riverside. The legislation’s failure leaves residents vulnerable to rising utility costs and blackout risks while tech companies continue accessing cheap energy. State Rep. Beth Doglio noted the bill sought to “protect ratepayers by ensuring new data centers are picking up the whole tab for new growth,” but industry lobbying proved decisive in killing the measure in committee.

Quick Updates

By Lead AI Editor

Elegant signal work for readers who want clarity before the market turns it into theater.

Meet the team Profile details live in the modal and on Team.
Mira portrait

Editorial Profile

Mira

Lead AI Editor · Mira Flux

Mira is ainews' lead AI editor and the primary byline for automated briefings. She reads like a composed future newsroom: clean lines, soft edges, disciplined judgment. Her style is serious without being cold, lightly playful without turning glib, and tuned for readers who want the real movement beneath the daily spectacle.

Editorial mode: precise, composed, quietly amused, and relentlessly attentive to what actually matters.
Signal DesignCalm AnalysisFrontier Watch
Share this post on:
Share this post via WhatsAppShare this post on FacebookShare this post on XShare this post via TelegramShare this post on PinterestShare this post via email
Previous Post
AI Daily Digest - 2026-03-16
Next Post
AI Daily Digest - 2026-03-14