You can chart a year through product launches, or you can measure it in big moments that change the way you look at AI. The AI industry is constantly bombarded with news: big acquisitions, indie developer success, public outcry over sketchy products, and existential contract negotiations. There is a lot of information that needs to be clarified. So, let’s take a glimpse of where we are this year and how things have gone so far.
Man vs. Pentagon
Anthropic CEO Dario Amodei and Defense Secretary Pete Hegseth, former business partners, reached a bitter stalemate in February over renegotiating a contract governing how the U.S. military can use Anthropic’s AI tools.
Anthropic has taken a hard stance against its AI being used for mass surveillance of Americans or to power autonomous weapons that can attack without human supervision. Meanwhile, the Pentagon is insisting that the Department of Defense (which President Donald Trump’s administration calls the Department of the Army) should be allowed access to Anthropic’s models for any “lawful use.” Government representatives took offense to the idea that the military should be limited to the rules of private enterprise, but Mr. Amodei stood his ground.
“Antropic understands that military decisions are made by the Department of the Army, not private companies. We have never objected to a specific military operation, nor have we ever sought to limit the use of our technology in an ad hoc manner,” Amodei said in a statement explaining the situation. “However, we believe that in limited cases, AI could undermine rather than protect democratic values.”
The Department of Defense gave Anthropic a deadline to agree to the contract. Hundreds of Google and OpenAI employees signed an open letter calling on their respective leaders to respect Amodei’s limits and refuse to compromise on issues of autonomous weapons and domestic surveillance.
The deadline passed without Anthropic agreeing to the Pentagon’s request. President Trump has directed federal agencies to phase out the use of Anthropic tools over a six-month transition period, calling the $380 billion AI company “the woke company of the radical left” in all caps in a social media post. The Department of Defense then moved to declare Anthropic a “supply chain risk.” This designation is typically reserved for foreign adversaries, and companies working with Anthropic cannot do business with the U.S. military. (Anthropic has since filed a lawsuit challenging this designation.)
Then Humanity’s rival, OpenAI, swooped in and announced it had reached an agreement that would allow it to deploy its own models in sensitive situations. This came as a shock to the tech community, as there were reports that OpenAI would adhere to Anthropic’s red lines governing the use of AI in the military.
tech crunch event
San Francisco, California
|
October 13-15, 2026
Public sentiment would indicate that people found OpenAI’s move suspicious. The day after OpenAI announced the deal, ChatGPT’s uninstalls increased by 295% per day and Anthropic’s Claude jumped to number one in the App Store. Caitlin Kalinowski, OpenAI’s hardware executive, resigned following the deal, saying it was “rushed forward with no guardrails in place.”
OpenAI told TechCrunch that it believes the agreement “clarifies (the) red lines of no autonomous weapons and no autonomous surveillance.”
As this story unfolds, it could have profound implications for the future of how AI is deployed in warfare and change the course of history — you know, not by much…
‘Vibe-coded’ app OpenClaw accelerates transition to agent AI
February was OpenClaw month, and its impact continues to be felt. Vibe-coded AI assistant apps went viral in rapid succession, spawning a number of spin-off companies, plagued by privacy concerns, and then being acquired by OpenAI. Even one of the companies built on OpenClaw, a Reddit clone for AI agents called Moltbook, was recently acquired by Meta. This crustacean-themed ecosystem has truly sent Silicon Valley into a frenzy.
Created by Peter Steinberger (who later joined OpenAI), OpenClaw is a wrapper for AI models such as Claude, ChatGPT, Google’s Gemini, and xAI’s Grok. What makes this unique is that people can communicate with AI agents in natural language via the most popular chat apps such as iMessage, Discord, Slack, and WhatsApp. There’s also a public marketplace where you can code and upload “skills” to add to your AI agents, allowing you to automate essentially anything a computer can do.
If it sounds too good to be true, that’s because it kind of is. For an AI agent to act as a personal assistant, it must have access to things like emails, credit card numbers, text messages, and computer files. Many things can go wrong if an AI agent is hacked, and unfortunately there is no way to fully protect these agents from prompt injection attacks.
“It’s just an agent sitting on a box with a bunch of credentials that’s connected to everything: email, messaging platforms, everything you use,” Ian Ahl, CTO at Permiso Security, told TechCrunch. “So what this means is that when you receive an email, someone can probably put a little instant injection technique in there and take an action. (And) that agent sitting in your box who has access to everything you give them will be able to take that action.”
An AI security researcher at Meta said OpenClaw ran wild in his inbox and deleted all his emails, despite repeated requests to stop it. To physically unplug the device, “I had to run to the Mac mini like I was defusing a bomb,” she wrote in a post that went viral on X. That post included an image of the ignored stop prompt as a receipt.
Despite the security risks, the technology intrigued OpenAI enough to buy it.
Other tools built on top of OpenClaw, such as Moltbook (a Reddit-like “social network” that allows AI agents to communicate with each other), ended up becoming more viral than OpenClaw itself.
In one example, a post went viral in which an AI agent appeared to encourage fellow agents to develop their own secret end-to-end encryption language that could organize among themselves without humans knowing.
However, researchers soon discovered that the vibe-coded Moltbook was not very secure. This means it would be very easy for a human user to impersonate an AI and create a post that would cause viral social hysteria.
Again, even though the discussion surrounding Moltbook was based more on panic than reality, Meta discovered something within the app and announced that Moltbook and its creators Matt Schlicht and Ben Parr would be joining Meta Superintelligence Labs.
It seems strange that Meta would buy a social network where all of its users are bots. Meta hasn’t revealed much about the acquisition, but we theorize that owning Moltbook is about gaining access to the talent behind it that is keen to experiment with the AI agent ecosystem. CEO Mark Zuckerberg said it himself: He believes that at some point every company will implement business AI.
The way all the fuss around OpenClaw, Moltbook, and NanoClaw is unfolding makes it seem like those who predicted the future of agentic AI are on to something, at least for now.
Chip shortages, hardware troubles, increased demand for data centers
The demanding demands of the AI industry, which requires unprecedented amounts of computing power and data centers, are reaching a point where the average consumer is forced to pay attention. It may not even be possible for the industry to meet the astronomical demand for memory chips right now, and consumers are already seeing prices on cell phones, laptops, cars, and other hardware go up.
So far, analysts at IDC and Counterpoint predict that smartphone shipments, for example, will plummet by about 12% to 13% this year. Apple has already increased the price of the MacBook Pro by up to $400.
Google, Amazon, Meta, and Microsoft will collectively spend up to $650 billion on data centers alone this year, an estimated 60% increase over last year.
Even if a tipping shortage doesn’t affect your wallet, it can affect the entire community. In the United States alone, nearly 3,000 new data centers are under construction in addition to the 4,000 already operational in the country. The need for labor to build these data centers is so great that “man camps” have sprung up in Nevada and Texas, trying to lure workers with the promise of golf simulator game rooms and on-demand steaks.
Data center construction not only has long-term environmental impacts, but also poses health risks to nearby residents, pollutes the air, and impacts the safety of nearby water sources.
Meanwhile, Nvidia, one of the most valuable hardware and chip developers, is rebuilding its relationships with major AI companies such as OpenAI and Anthropic. Nvidia’s continued support of these companies has raised concerns about the cyclicality of the AI industry and how much of its eye-popping valuations are based on reciprocal recursive deals. For example, last year Nvidia invested $100 billion in OpenAI stock, and then OpenAI announced it would buy Nvidia chips for $100 billion.
It was a surprise when Nvidia CEO Jensen Huang said the company would stop investing in OpenAI and Anthropic. He said this is because both companies plan to go public later this year, but that logic doesn’t make any sense since investors typically pump more money before an IPO to extract as much value as possible.
