Google rewrites headlines in Search & National U.S. AI law push - Tech News (Mar 22, 2026)

Let’s start with the fight over how news is presented online. Google Search is experimenting with replacing publishers’ original headlines in standard search results with AI-generated alternatives. The surprising part isn’t just that titles get shortened—that’s been common for years—but that these are rewritten in new words, sometimes shifting the tone or even the apparent stance of the article. Google says it’s a small, narrow test aimed at relevance, and not specifically focused on news. Still, it puts a spotlight on a big question: if platforms can rewrite the label on journalism, who’s responsible when that label becomes misleading? For publishers, headlines are not decoration—they’re part of accuracy and accountability.

Staying on the theme of AI and control, the Trump administration has released a legislative framework for a single national AI policy—one that would set federal safety and security rules while blocking states from making their own AI regulations. The outline spans topics from child-safety protections to the practical realities of AI infrastructure like data centers and energy use, and it gestures toward federal guidance on AI-related intellectual-property disputes. It also calls for rules meant to prevent AI from being used to silence lawful political speech. The political hook here is the preemption: Washington would take the wheel, and states like New York and California would be told to stand down. The administration argues that a unified standard beats a patchwork, and major AI companies have been lobbying for exactly that. But turning it into law won’t be simple in a narrowly divided Congress with competing priorities—and the end result could shape how AI is built and deployed across the entire U.S. economy.

Now to the darker edge of AI in practice: the US- and Israel-led war with Iran is being described as an early large-scale test of “AI-enabled” warfare. Reports point to an unusually fast tempo of strikes—hundreds of targets hit in a day—where AI tools reportedly help sift intelligence quickly enough to compress the time between identifying something and acting on it. The Pentagon continues to insist humans make the final strike decisions, but speed itself changes the accountability landscape. This is also surfacing tension between defense contractors, AI firms, and government policy—especially around guardrails, surveillance concerns, and what kinds of autonomy should be off-limits. And with an investigation underway after a strike reportedly hit a girls’ school, the central issue becomes painfully concrete: when systems accelerate decisions at scale, it can be harder to untangle what went wrong, who knew what, and how to prevent it next time.

A related story out of Ukraine shows how fast innovation is moving at the front lines. Ukrainian units are testing and refining homebuilt interceptor drones designed to shoot down Shahed loitering munitions—drones Russia launches in large waves. Early in the war, Ukraine had few good options against this kind of threat, and expensive missile defenses don’t scale well when the attacker’s goal is to overwhelm you with volume. Now, frontline feedback is driving rapid design changes, with soldiers, local manufacturers, and volunteer networks iterating in near real time. The wider significance is that this is becoming a template: cheaper, adaptable air defense that can evolve quickly, and that other countries are watching closely as the same Iranian-designed drones appear in more places.

Switching from battlefields to personal safety online, Germany’s Justice Ministry says it’s preparing legislation to criminalize the creation and distribution of pornographic deepfakes. The goal is to close the gap between laws that address physical-world sexual abuse and the newer reality of digital impersonation—where someone can be turned into explicit content without consent, at scale. The draft would also expand police authority to search suspects’ devices and add civil-law tools to help victims pressure platforms, including efforts to identify perpetrators and suspend accounts. A high-profile complaint from actress Collien Fernandes helped accelerate attention, but officials are also candid about the hard part: enforcement, especially when offenders and infrastructure cross borders. The message, though, is clear—this is being treated as a form of sexual violence, not a prank or a niche internet issue.

On the creator economy side, WordPress.com is rolling out AI-agent capabilities that can do more than suggest text. These connected assistants can draft and edit posts, help publish content, and even handle maintenance chores like moderating comments or reorganizing categories—based on natural-language instructions from the site owner. WordPress says actions are logged and approvals are required, with AI-generated posts defaulting to drafts unless a user decides otherwise. Why it’s interesting is the scale: WordPress powers a huge slice of the web, so this isn’t just a convenience feature for a few teams. It could lower the barrier to publishing even further—while also accelerating the flood of machine-generated pages, and pushing the web into a new phase of questions about quality, authenticity, and what “human-made” should mean online.

Looking up—literally—NASA says it’s ready to launch Artemis 2 on April 1, the first crewed lunar mission in about 50 years. The mission echoes Apollo-era ambition, but the context has changed. The old rivalry with Russia isn’t the driver; China is the more relevant competitor now, and modern missions tend to prioritize validation and safety over spectacle. Artemis 2 is set up to test Orion’s systems thoroughly and rehearse maneuvers needed for future missions, including the choreography required to eventually rendezvous with a lunar lander. NASA is also leaning hard into biomedical monitoring to understand how deep-space conditions affect human health. In other words, this flight is as much a stress test and a science mission as it is a symbolic return to deep-space crewed exploration.

Finally, a medical breakthrough with big implications for biotech and immunotherapy. Researchers at Mass General Brigham and Dana-Farber report that a single injection of a genetically engineered oncolytic virus may help make glioblastoma—one of the most treatment-resistant brain cancers—more vulnerable to immune attack. In a phase 1 trial involving patients with recurrent disease, the treatment was associated with survival that looked better than historical expectations, and tissue analyses suggested it drew killer T cells into tumors and kept them active. If this direction holds up in larger trials, it offers something the field badly needs: a credible strategy to make an immune “cold” tumor behave more like one the immune system can actually recognize and fight. For a cancer where standard care has barely budged for decades, even early signs like this get a lot of attention.