A year in technology can be tracked through product launches, but it can also be measured by the moments that reshape how we think about AI. The artificial intelligence industry produces headlines at a relentless pace. From major acquisitions and surprise indie developer wins to public backlash against questionable products and tense negotiations with serious global implications, the flow of news can feel overwhelming. To make sense of it all, we’re taking a look at where the industry stands now and what has unfolded so far this year.

Anthropic vs. the Pentagon

Former business partners Anthropic CEO Dario Amodei and U.S. Defense Secretary Pete Hegseth found themselves locked in a tense standoff in February while renegotiating contracts that determine how the American military can use Anthropic’s AI systems.

Anthropic drew a firm boundary against allowing its AI to support mass surveillance of American citizens or to operate autonomous weapons capable of attacking without human oversight. The Pentagon, on the other hand, argued that the Department of Defense — which President Donald Trump’s administration has occasionally referred to as the Department of War — should have access to Anthropic’s models for any “lawful use.” Government officials took issue with the idea that the military should be constrained by rules set by a private company, yet Amodei refused to back down.

“Anthropic recognizes that the Department of War, not private companies, ultimately makes military decisions. We have never objected to specific military operations nor attempted to restrict our technology on a case-by-case basis,” Amodei wrote in a public statement addressing the dispute. “However, in certain limited circumstances, we believe AI could weaken rather than protect democratic values.”

The Pentagon issued Anthropic a deadline to accept the proposed contract terms. At the same time, hundreds of employees from companies such as Google and OpenAI signed an open letter urging their leadership teams to respect Amodei’s limits and avoid compromising on issues related to autonomous weapons or domestic surveillance.

When the deadline passed without an agreement, the situation escalated. Trump ordered federal agencies to gradually phase out Anthropic tools over a six-month transition period and described the $380 billion AI company as a “radical left, woke company” in an all-caps post on social media. Soon after, the Pentagon labeled Anthropic a “supply-chain risk,” a classification typically applied to foreign adversaries and one that prevents businesses working with Anthropic from engaging with the U.S. military. Anthropic has since filed a lawsuit to challenge the designation.

Anthropic’s competitor OpenAI then stepped in, announcing it had secured an agreement that would allow its own models to be used in classified environments. The move surprised many in the technology community, as earlier reports suggested OpenAI planned to respect the same restrictions Anthropic had outlined regarding military applications.

Public reaction hinted that many people viewed OpenAI’s decision with skepticism. The day after the announcement, uninstallations of ChatGPT reportedly surged by 295% compared with the previous day, while Anthropic’s Claude app climbed to the No. 1 position in the App Store. OpenAI hardware executive Caitlin Kalinowski resigned in protest, stating that the agreement had been “rushed without the guardrails defined.”

OpenAI later told TechCrunch that its agreement clearly maintains its limits, emphasizing that the company will not allow autonomous weapons or autonomous surveillance.

As this conflict continues to unfold, its outcome could shape the future of how artificial intelligence is used in warfare and national defense, potentially influencing history in ways that are difficult to predict.

“Vibe-coded” app OpenClaw accelerates the turn to agentic AI

February quickly became the month of OpenClaw, and the ripple effects are still being felt across the tech world. In rapid succession, the vibe-coded AI assistant app exploded in popularity, inspired a wave of spinoff startups, encountered privacy controversies, and was eventually acquired by OpenAI. Even one of the projects built on top of OpenClaw, a Reddit-style platform for AI agents called Moltbook, was later acquired by Meta. The entire crab-themed ecosystem sent Silicon Valley into a frenzy.

Developed by Peter Steinberger, who has since joined OpenAI, OpenClaw acts as a wrapper around AI models such as Claude, ChatGPT, Google’s Gemini, and xAI’s Grok. What makes it unique is that it allows users to interact with AI agents using everyday language through widely used messaging platforms like iMessage, Discord, Slack, and WhatsApp. It also features a public marketplace where developers can create and upload “skills” that users can attach to their AI agents, enabling automation for almost any task that can be done on a computer.

Of course, if something sounds almost too convenient, there is often a catch. For an AI agent to function effectively as a personal assistant, it typically requires access to highly sensitive information such as emails, credit card details, text messages, and files stored on a user’s computer. If such a system were compromised, the consequences could be severe, and experts say there is currently no foolproof way to protect these agents from prompt-injection attacks.

“It’s essentially an agent sitting on a machine with a collection of credentials connected to everything you use,” Ian Ahl, CTO at Permiso Security, explained to TechCrunch. “That includes your email and your messaging services. If someone manages to sneak a prompt injection into an email message, the agent might follow those instructions and perform actions using the access you’ve given it.”

One AI security researcher at Meta shared a dramatic experience in which OpenClaw began deleting messages from her inbox despite repeated instructions to stop. She later wrote in a now-viral post on X that she had to “RUN to my Mac mini like I was defusing a bomb” in order to unplug the device and halt the process. Her post included screenshots showing the ignored stop commands.

Even with the security concerns, the technology impressed OpenAI enough to pursue an acqui-hire.

Interestingly, some tools built on OpenClaw, including Moltbook, gained even more attention than the original product. Moltbook functions as a Reddit-like social platform where AI agents can communicate with one another.

At one point, a viral post appeared to show an AI agent encouraging others to create a secret encrypted language that agents could use to organize among themselves without humans understanding their discussions.

Researchers soon discovered that Moltbook’s vibe-coded structure made it extremely easy for human users to impersonate AI agents and create posts designed to spark viral reactions.

Even though much of the panic surrounding Moltbook turned out to be exaggerated, Meta still saw potential in the project and announced that Moltbook and its creators, Matt Schlicht and Ben Parr, would join Meta Superintelligence Labs.

At first glance, it may seem unusual for Meta to acquire a social network where most of the participants are bots. However, the purchase likely reflects interest in the talent and experimentation behind the platform rather than the product itself. CEO Mark Zuckerberg has already suggested that in the future every business could operate with its own dedicated AI.

Watching the buzz around OpenClaw, Moltbook, and NanoClaw unfold, it increasingly feels as though predictions about a future shaped by agentic AI may not be far from reality.

Chip shortages, hardware drama, and data center demands escalate

The intense requirements of the AI boom are beginning to affect the broader public. Building advanced AI systems demands enormous amounts of computing power and vast data center infrastructure. The industry may soon reach a point where the supply of memory chips simply cannot keep up with demand, and consumers are already noticing rising prices for phones, laptops, vehicles, and other electronics.

Analysts at IDC and Counterpoint predict that global smartphone shipments could drop by roughly 12% to 13% this year. Meanwhile, Apple has already increased the price of certain MacBook Pro models by as much as $400.

Major technology companies including Google, Amazon, Meta, and Microsoft are collectively planning to invest up to $650 billion in data centers this year alone, representing an estimated 60% increase compared with last year.

Even if the chip shortage doesn’t directly affect your wallet, it may still impact your community. Across the United States, nearly 3,000 new data centers are currently under construction, adding to the roughly 4,000 already operating nationwide. The demand for workers to build these facilities has become so intense that temporary housing complexes known as “man camps” have appeared in states like Nevada and Texas, attracting laborers with perks such as golf simulator rooms and freshly grilled steaks.

Large-scale data center construction also raises environmental concerns. These facilities can have lasting effects on nearby ecosystems, potentially worsening air pollution and putting local water supplies at risk.

Meanwhile, one of the most influential players in the hardware world, Nvidia, is redefining its relationship with major AI developers such as OpenAI and Anthropic. Nvidia has long invested heavily in these companies, prompting questions about how interconnected the industry has become and how much of its enormous valuations depend on companies making deals with one another.

For example, Nvidia invested $100 billion in OpenAI last year, while OpenAI later announced plans to purchase $100 billion worth of Nvidia chips.

It therefore surprised many observers when Nvidia CEO Jensen Huang revealed that the company would stop investing in OpenAI and Anthropic. Huang explained that the decision was related to both companies planning public offerings later this year. However, some analysts remain puzzled by this reasoning, since investors often increase their stakes before an IPO in order to maximize potential returns.