US bans Anthropic over AI limits & China’s new five-year tech push - Tech News (Mar 15, 2026)

We’ll start with that U.S. government and AI clash, because it’s unusually direct. The Trump administration has ordered the termination of federal contracts with Anthropic after the company’s CEO, Dario Amodei, wouldn’t remove usage restrictions that block Claude from being used for mass surveillance or fully autonomous weapons without human oversight. Anthropic is suing, calling it unlawful retaliation and warning that branding it a “supply chain risk” could pressure other defense contractors to drop it too. What makes this case especially consequential is the alleged dependency: Claude is described as deeply embedded in classified workflows and in systems used for intelligence analysis. If that’s accurate, this isn’t just a values fight—it’s also an operational headache. Zooming out, the bigger story is that the rules for military and surveillance AI are still mostly vibes and internal policies, not clear law. And when the law is missing, power struggles like this one tend to decide the boundaries.

Staying with geopolitical tech competition, China has now approved and published its 15th five-year plan for 2026 to 2030—and it’s a statement of intent. The plan calls for what it describes as extraordinary measures to make China a global leader in artificial intelligence, quantum technology, and other frontier fields. What’s notable is the tone shift: it’s less about catching up with the United States and more about trying to set the pace. Beijing is framing science and R&D as core national priorities alongside defense, economic growth, and international influence. And it leans hard into technological self-reliance—explicitly naming bottlenecks like advanced chips, industrial tools, high-end instruments, foundational software, and advanced materials. The plan also expands the “AI plus” campaign—basically, embedding AI across industry and governance—and treats AI supply chains like a strategic security issue. The underlying message is simple: in a world where access can be cut off, control of the stack becomes a national strategy.

Another China story also shows how fast that strategic push is moving into real-world products. China has approved a brain implant for commercial use aimed at helping some people with paralysis regain hand function—described as the first commercial authorization of its kind. The system reads brain activity linked to the intention to move, then drives a robotic glove to open and close, enabling basic grasping. This is not sci-fi mind control, and it won’t be for everyone—but it’s a meaningful step toward turning brain-computer interfaces into an actual category, not just lab demos. It also raises the stakes in global neurotech competition. With Neuralink aiming for higher-volume production next year, and multiple research groups pushing similar concepts, we’re heading into an era where the big questions won’t only be medical. They’ll be about regulation, privacy of neural data, and who gets access first.

Now to the information battlefield—because the Iran–U.S. conflict is generating a wave of AI-generated deepfake videos that researchers say is bigger than what we’ve seen in past wars. On X, fabricated clips showing everything from captured soldiers to burning embassies are circulating alongside real footage, making it harder for everyday users to tell what’s authentic. X says it’s updating enforcement: accounts in its revenue-sharing program can be temporarily suspended for posting AI-made war content without disclosure, with repeat offenses leading to permanent suspensions. That’s a clear acknowledgment that monetization can fuel misinformation. But monitors say the platform is still flooded, and that some high-reach premium accounts keep pushing viral fakes faster than corrections can spread. Even more awkward: fact-checkers found that X’s chatbot, Grok, sometimes validated fake visuals as real—adding a machine-made stamp of confidence to the confusion. The broader issue here isn’t just deepfakes; it’s how quickly the attention economy turns uncertainty into engagement.

That tension between safety, trust, and platform design shows up again in Meta’s latest move: Instagram says it will discontinue end-to-end encrypted Direct Messages starting May 8th, 2026. End-to-end encryption meant only the people in a chat could read the messages. Rolling it back means Meta can technically access contents when required—and Meta argues this change will enable stronger AI-driven detection of harmful and illegal material, including child exploitation content, grooming, scams, and harassment. This sits right in the middle of a growing global policy tug-of-war. Governments want platforms to find illegal content even inside private messaging. Privacy advocates argue that weakening encryption increases risk for everyone, especially journalists, activists, and abuse victims. Regardless of where you land, the significance is that “private by default” messaging is no longer a one-way trend. Platforms are increasingly being pulled toward inspectable systems—either by regulators, or by the incentives of automated moderation.

Next, a major international cybercrime disruption: Interpol says an operation called Synergia III sinkholed roughly forty-five thousand IP addresses and seized servers tied to cybercrime worldwide. The effort ran across dozens of countries and led to arrests and additional suspects under investigation. Interpol highlighted everything from social-engineering fraud—like romance scams and sextortion—to phishing networks impersonating banks, government portals, and payment services. The practical takeaway is that cybercrime at scale relies on infrastructure: servers, domains, and command systems. When law enforcement coordinates internationally, it can meaningfully raise the cost for criminals—at least temporarily. The less comforting takeaway is that the volume is so large that even a big operation like this is more like pruning than eradication.

Quickly, something more uplifting from space science: researchers now say NASA’s DART mission didn’t just change the orbit of the small asteroid moon it hit—it also measurably changed the pair’s trajectory around the sun. After the 2022 impact, DART was celebrated for shortening the moon’s orbit around its partner asteroid. But detecting a change in the system’s solar path required years of follow-up, combining extremely precise tracking methods. The newly reported result is subtle—think a tiny shift that adds up over time—but it’s the first direct observation of a kinetic impact changing an asteroid system’s heliocentric orbit. Even more interesting, debris blasted off the asteroid added momentum, effectively amplifying the push. For planetary defense, this matters because the difference between “barely moved” and “moved enough” can hinge on how the target responds, not just the size of the spacecraft.

Finally, a story that sits at the intersection of biotech and the maker mentality: an Australian entrepreneur worked with researchers to create an experimental personalized mRNA cancer vaccine for his rescue dog, Rosie—using AI tools to help process genetic data and design a target. After treatment, one tumor reportedly shrank significantly, improving comfort, though the cancer hasn’t disappeared and another tumor didn’t respond. Researchers called the result surprising, but scientists also warned that this kind of single case can get overhyped without controlled studies. Why it’s still worth noting: it hints at where medicine is heading—toward faster, more tailored therapies. And pets can sometimes become early proving grounds for approaches that later inform human treatments, especially as sequencing and design pipelines speed up.