## Trump Administration Keeps Door Open for Escalating Action Against Anthropic

The Trump administration has refused to rule out additional punitive measures against artificial intelligence company Anthropic, signaling that the ongoing conflict between the AI startup and the federal government is far from over.

During a videoconference hearing on Tuesday, Justice Department attorney James Harlow bluntly told US District Judge Rita Lin that he was "not prepared to offer any commitments" on whether the government would refrain from further action against the company. The statement came as reports emerged that President Trump is in the final stages of drafting an executive order that would formally prohibit federal agencies from using any of Anthropic's products — a move that was first reported by Axios and confirmed by a White House source not authorized to speak publicly on the matter.

## Lawsuits Filed After Government Labels Anthropic a Security Risk

Tuesday's court appearance grew out of one of two federal lawsuits Anthropic filed against the Trump administration on Monday. The company alleges that the government unconstitutionally designated it a supply-chain risk, effectively turning it into an outcast within the technology industry. According to Anthropic, the consequences have been severe — billions of dollars in potential revenue are now in jeopardy as existing clients walk away from deals and prospective customers demand renegotiated terms.

The company is now seeking a preliminary court order to suspend the risk designation and prevent the administration from taking any further retaliatory steps.

### Court Hearing Scheduled Amid Mounting Business Losses

Tuesday's session focused on determining the timeline for a full preliminary hearing. Anthropic's legal team, represented by WilmerHale attorney Michael Mongan, expressed urgency in moving quickly to limit ongoing damage to the company's operations. Mongan indicated he would accept an April hearing date only if the administration agreed to pause any additional action — a commitment the government refused to make.

"The actions of defendants are causing irreparable injuries, and those injuries are mounting day by day," Mongan told the court.

Following the government's refusal, Judge Lin moved the hearing date to March 24 in San Francisco — earlier than originally planned but still later than Anthropic had hoped. The judge acknowledged the high stakes on both sides, stating she wanted to rule on a complete record despite the expedited circumstances.

A parallel lawsuit filed in Washington, DC, remains on hold while Anthropic pursues an administrative appeal with the Department of Defense — a process widely expected to be unsuccessful.

## The Root of the Dispute: Military Use of AI Technology

The conflict between Anthropic and the Pentagon has been building for months. At its core, the dispute centers on Anthropic's refusal to authorize the military's unrestricted use of its AI technology for any lawful purpose. The company expressed concern that such broad authorization could enable mass surveillance of American citizens or allow the launching of missiles without direct human oversight. The Defense Department, however, maintains that decisions about how it deploys technology fall entirely within its own authority.

## Legal Experts Say Administration Is Overstepping

A number of legal scholars with backgrounds in constitutional law and government contracting argue that the administration's actions against Anthropic follow a troubling pattern of weaponizing legal mechanisms to punish entities perceived as politically unfriendly — a list that has grown to include universities, media organizations, and prominent law firms.

Harold Hongju Koh, a Yale Law School professor who previously served in the Obama administration, sees the Anthropic case as part of a broader and unmistakable trend.

"If this is a one-off, you might give the president some deference," Koh said. "But now, it's just unmistakable that this is the latest in a chain of events related to a punitive presidency."

David Super, a constitutional law professor at Georgetown University Law Center, took direct aim at the Pentagon's legal reasoning, arguing that the provisions used to sanction Anthropic were originally crafted to guard against enemy sabotage — not to discipline contractors who refuse to comply with every government demand.

"It is an absurd stretch of the English language to equate 'does not agree to every demand of Pete Hegseth' with 'sabotage,'" Super said, referring to the secretary of defense. He also noted that the Supreme Court has consistently pushed back against this kind of legal overreach, citing recent rulings against presidential tariffs and earlier decisions invalidating actions by the Biden administration.

## Tech Industry Left in Uncertainty

While the legal battles play out, companies that depend on Anthropic's Claude suite of AI tools are wrestling with difficult practical questions about whether to seek alternative solutions. Meanwhile, rival AI giants OpenAI and Google are reportedly moving ahead with their own Pentagon contracts, stepping into the void left by Anthropic — despite internal pressure from their own employees to resist government demands over technology use.

Zohra Tejani, a partner at law firm Seyfarth Shaw who advises technology companies navigating federal contracts, believes Anthropic may ultimately succeed in having the supply-chain-risk label removed and could eventually restore many of its business relationships. However, she cautioned that winning back the trust and contracts of the current administration would be a far steeper climb.

## A Warning Shot to the Entire AI Industry

Beyond Anthropic's immediate situation, legal and industry experts warn that the government's aggressive stance could have a chilling effect across the broader AI sector. Even if the courts ultimately side with Anthropic, the damage done to its reputation and revenue — combined with the resources required to fight the legal battle — may deter other technology companies from standing up to government pressure.

"The Pentagon is sending a message to every other AI company: If you defy the Pentagon, you risk nationalization and heavy-handed government intervention," said Christoph Mlinarchik, a former Pentagon contracting officer who now advises federal suppliers. "The Pentagon does not want to cede veto or moral authority to contractors, no matter the flavor of technology."

For the Trump administration, this outcome — a tech industry walking on eggshells — may represent a strategic victory in its own right, regardless of how the courts eventually rule.

The Trump administration won't commit to halting action against Anthropic as reports emerge of an executive order banning the AI firm from all federal agencies.

Trump Refuses to Rule Out Further Action Against Anthropic

large_Trump-Admin-Refuses-to-Say-Wont-Take-Further-Action-Against-Anthropic-Business-2265661562.jpg

medium_Trump-Admin-Refuses-to-Say-Wont-Take-Further-Action-Against-Anthropic-Business-2265661562.jpg

small_Trump-Admin-Refuses-to-Say-Wont-Take-Further-Action-Against-Anthropic-Business-2265661562.jpg

thumbnail_Trump-Admin-Refuses-to-Say-Wont-Take-Further-Action-Against-Anthropic-Business-2265661562.jpg

Trump-Admin-Refuses-to-Say-Wont-Take-Further-Action-Against-Anthropic-Business-2265661562.jpg

Trump Administration Refuses to Rule Out Further Crackdown on AI Company Anthropic

## Interstellar Comet 3I/Atlas Stuns Scientists With Unusually High Alcohol Content

A remarkable discovery has emerged from one of the world's most powerful radio observatories: the interstellar comet 3I/Atlas is packed with methanol — a form of alcohol widely used in fuels and industrial solvents — at concentrations far exceeding anything typically seen in comets native to our solar system.

### What the ALMA Telescope Revealed

Data collected by the ALMA telescope, situated in the high-altitude Atacama Desert of Chile, shows that the coma surrounding 3I/Atlas is unusually rich in methanol. While this organic compound does appear in solar system comets, the levels detected in 3I/Atlas are up to four times higher than the standard amount observed elsewhere.

According to a study published on arXiv, this makes 3I/Atlas the second most methanol-rich comet ever documented. The only comet to surpass it is C/2016 R2, an already anomalous object first identified a decade ago. Alongside methanol, separate research efforts have also identified elevated concentrations of carbon dioxide, iron, and nitrogen within the comet — further underlining just how chemically distinctive this visitor truly is.

### A Chemistry Unlike Anything in Our Solar System

The unusual chemical profile of 3I/Atlas has led researchers to propose that this comet originated in a radically different environment from anything found in our own cosmic neighborhood. The excess methanol, the carbon dioxide-dominated coma, and the atypical ratios of various compounds collectively point toward a birthplace that was either significantly colder, subjected to more intense radiation, or governed by entirely different chemical processes than those that shaped comets in our solar system.

This makes 3I/Atlas not just a scientific curiosity, but a potential window into the diverse planetary-forming conditions that exist across the galaxy.

### Could 3I/Atlas Be a Hyperactive Comet?

The research team has also raised the possibility that 3I/Atlas belongs to a special class of bodies known as hyperactive comets. These are objects that emit more water vapor than their surface area alone could reasonably produce. In such comets, gaseous outgassing does not come solely from the nucleus itself — a portion is also generated by tiny ice grains drifting through the coma, which sublimate as the comet warms.

Scientists suggest that in the case of 3I/Atlas, both methanol and water vapor were released simultaneously from the solid nucleus and from these free-floating icy particles. A notable share of the methanol detected appears to have originated from detached ice deposits that began sublimating as the comet drew closer to the Sun — behavior entirely consistent with known hyperactive comet dynamics.

### Ruling Out an Artificial Origin

This pattern of activity further strengthens the scientific consensus that 3I/Atlas is a completely natural object — one that is extraordinarily cold and chemically complex, but natural nonetheless. Any lingering speculation about an artificial or manufactured origin has been effectively dismissed by these findings.

### A Rare Interstellar Traveler Heading Back Into the Void

At present, 3I/Atlas is departing our solar system at a speed of approximately 60 kilometers per second, never to return. It holds the distinction of being only the third confirmed interstellar object ever observed by humanity. Despite its rarity, astronomers are optimistic that upcoming generations of more advanced telescopes and detection systems will uncover many more such interstellar travelers in the years ahead, gradually expanding our understanding of the universe beyond our own cosmic backyard.

Scientists using the ALMA telescope discover that interstellar comet 3I/Atlas contains methanol levels up to four times higher than typical solar system comets.

Interstellar Comet 3I/Atlas Found Rich in Methanol Alcohol

large_GettyImages-183097875.jpg

medium_GettyImages-183097875.jpg

small_GettyImages-183097875.jpg

thumbnail_GettyImages-183097875.jpg

GettyImages-183097875.jpg

Interstellar Comet 3I/Atlas Stuns Scientists With Unusually High Alcohol Content

## GPS Under Attack: What's Happening Near Iran

Across conflict zones in and around Iran, a hidden electronic battle is unfolding — one that most civilians never see but increasingly feel. Delivery drivers arrive at wrong addresses, mapping apps spin in circles, and aircraft systems silently display incorrect positions. The culprit behind this growing chaos is electronic warfare targeting satellite navigation systems.

Disrupting GPS has become a standard military tactic in modern conflicts. By interfering with satellite signals, armed forces can throw off enemy drones, precision missiles, and surveillance equipment. The problem is that civilian technology runs on exactly the same signals. When those signals are compromised, the fallout extends far beyond the battlefield — hitting commercial airlines, global shipping networks, logistics operations, and the location-based apps built into our daily routines.

## Two Distinct Methods: Jamming vs. Spoofing

GPS disruptions typically occur through one of two techniques — jamming and spoofing — and understanding the difference is key to grasping why these attacks are so difficult to detect and counter.

### GPS Jamming: Drowning Out the Signal

GPS satellites orbit roughly 12,400 miles above Earth and transmit signals at around 50 watts of power. By the time that signal reaches the ground, it has weakened considerably — making it surprisingly vulnerable to interference.

Jamming works by blasting a stronger noise signal that overwhelms the faint GPS transmission. Jim Stroup, head of growth for SandboxAQ's navigation product AQNav, describes it this way: "It's like saturating out your eyeball: you're trying to see something really far away, and someone comes by with a flashlight, and now you can't make sense of it." Even a cheap, battery-powered jammer purchased online can knock out GPS reception across a wide local area.

### GPS Spoofing: Faking the Signal

Spoofing is considerably more sophisticated — and far more dangerous. Rather than simply blocking GPS signals, a spoofer impersonates them. It captures real satellite transmissions and rebroadcasts convincing fake versions, tricking receivers into calculating a completely wrong position.

The unsettling part? Everything appears to be working fine. The navigation display looks normal. No alarms go off. The system just shows the wrong location.

A spoofing device listens to legitimate satellite signals, then rebroadcasts slightly altered versions. The target device — whether a drone, a ship, or a commercial aircraft — picks up this fraudulent signal as if a new satellite had appeared. As the receiver incorporates the fake data, it gradually drifts off course without any indication that something is wrong.

"You can actually take a drone and steer it off course. And to the drone and to the pilots, everything on GPS will look like it's operationally just fine," Stroup explains. The implications are serious: a bad actor could manipulate a drone into crossing an international border, potentially triggering a geopolitical incident.

## The Ripple Effect: Far Beyond Broken Map Apps

For most people, a GPS glitch might mean a late food delivery or a frustrating detour. But the consequences of sustained GPS disruption stretch into critical infrastructure in ways that few people consider.

Hospitals, power utilities, and even nuclear facilities depend on GPS — not primarily for navigation, but for precise timing. Dozens of highly sensitive systems within a single facility must operate in perfect synchronization. GPS provides the timing backbone that keeps them aligned.

"It's not so much that they just need to know what time it is," Stroup says. "It's the fact that they have 18 disparate, highly sensitive technical systems that need to run on Swiss-like precision. If there's one thing that's slightly out of alignment, that can cause catastrophic issues."

A large-scale or prolonged GPS disruption could mean grounded commercial flights, strained power grids, and health care facilities struggling to maintain synchronized safety systems — all simultaneously.

## Can We Build a Better Navigation System?

Experts refer to GPS alternatives as alternate PNT — position, navigation, and timing solutions. However, no single alternative currently addresses all three functions at once. Some handle positioning and navigation but not timing. Others manage timing but fall short on location accuracy.

### Visual Navigation: A High-Tech Old Trick

One approach, known as visual navigation (vis-nav), is essentially a modernized version of what pilots did before GPS existed. Early aviators would look down, identify a recognizable landmark, and use it to confirm their position. Today, onboard cameras and computers can perform the same task far more rapidly.

The limitation is significant, however. Visual navigation depends on recognizable, stable terrain features. It performs well on fixed routes but collapses over open oceans, polar regions, or war zones where infrastructure and landmarks have been destroyed.

### Low-Earth-Orbit Satellites: A Partial Answer

Some proposed solutions involve repurposing signals from networks like Starlink or building denser constellations of low-Earth-orbit satellites. The problem, as Stroup notes, is straightforward: "They can just be jammed as well. Maybe you can't go buy one off eBay with a 9-volt battery — maybe now you need a car battery."

More power, yes. A real solution, not necessarily.

### Quantum Magnetic Navigation: The Earth as a GPS Substitute

Perhaps the most promising long-term solution involves using the Earth itself as a navigation reference. Buried mineral deposits, rock formations, and mountain ranges each create subtle, measurable disturbances in the planet's magnetic field. No two locations on Earth share exactly the same magnetic signature.

Using next-generation quantum sensors, it is now possible to detect these minute fluctuations and construct what researchers call a "dynamic anomaly map" — essentially a unique magnetic fingerprint for every point on the globe. A vehicle equipped with these sensors can match its real-time magnetic readings against the map to determine its precise location.

The critical advantage: this signal originates from within the Earth itself. It cannot be jammed or spoofed the way satellite signals can. "It doesn't have the same susceptibility to jamming or interference you would have with the satellite constellations," Stroup says.

His recommendation is not to abandon GPS entirely but to layer these complementary technologies on top of it, creating a resilient, redundant navigation ecosystem that is far harder to disrupt.

## The Bigger Picture

GPS vulnerabilities near Iran and other conflict zones are a reminder that much of the world's critical infrastructure rests on a surprisingly fragile foundation. As electronic warfare becomes more sophisticated, the case for investing in layered, interference-resistant navigation and timing systems grows stronger — not just for military purposes, but for the hospitals, power grids, airlines, and delivery apps that billions of people depend on every day.

GPS jamming and spoofing near Iran are disrupting airlines, logistics, and navigation apps. Learn how these attacks work and what solutions exist.

GPS Attacks Near Iran Disrupt Flights and Delivery Apps

large_Navigation_LEAD.gif

medium_Navigation_LEAD.gif

small_Navigation_LEAD.gif

thumbnail_Navigation_LEAD.gif

Navigation_LEAD.gif

How GPS Attacks Near Iran Are Disrupting Flights, Deliveries, and Everyday Navigation

## AI Disinformation Runs Rampant on X Amid Iran Conflict

Since the United States and Israel launched military operations against Iran on February 28, social media platform X has been overwhelmed by a torrent of disinformation — much of it powered by artificial intelligence. What began as a flood of misleading content has since evolved into a sophisticated, AI-fueled misinformation campaign that is making it increasingly difficult for users to distinguish fact from fiction.

### Grok Gets It Wrong — Then Makes It Worse

X's own AI chatbot, Grok, has been at the center of the problem. The tool repeatedly misidentified the location and timestamp of a video originally posted by an Iranian state-owned media outlet, then compounded the error by generating an AI image in an apparent attempt to support its incorrect claims. Independent conflict analyst Hagin, who was verifying content on the platform, called the chatbot's behavior "AI slop of destruction" — a telling sign of just how far removed from reality the platform has drifted.

## The Scale of Fake AI Content on X

The volume and sophistication of AI-generated disinformation has surged in recent days. Fabricated videos and images are being circulated by paid accounts holding blue check marks, as well as by Iranian officials attempting to portray inflated accounts of wartime destruction.

Some notable examples include:

- **AI-generated footage** of a high-rise building in Bahrain engulfed in flames, shared by Iranian state media on March 2.
- **A fabricated image** of a U.S. B-2 stealth bomber being shot down by Iran, with American troops shown as prisoners — viewed more than one million times before deletion.
- **Images depicting Delta Force soldiers** allegedly captured by Iranian authorities, which racked up over five million views before being removed.
- **A cave-based missile manufacturing video**, widely recognized as unrealistic but still viewed more than one million times across multiple accounts.

### Antisemitic Narratives Pushed Through AI Content

Researchers at the Institute for Strategic Dialogue (ISD) have identified a coordinated pro-regime propaganda network on X that is leveraging AI-generated posts to spread overtly antisemitic content. These posts depict Orthodox Jewish individuals leading American soldiers into battle or celebrating U.S. casualties — a deliberate effort to weaponize AI imagery for ideological purposes.

Accounts within the same network also shared a fabricated video falsely showing young girls walking past President Donald Trump in a state of undress. The post garnered an estimated 6.8 million views before being taken down, though it continues to circulate on the platform.

## A Turning Point for AI-Fueled Fake News

"What is particularly unique about this war is the dramatic uptick in AI-generated content I find myself debunking," Hagin told WIRED. "This is likely due to AI being advanced enough to fool journalists, and the ease with which users can create this AI slop with zero consequences. The longer we go without regulations against AI abuse, the more harm will be caused."

The accessibility of modern AI image and video generation tools has dramatically lowered the barrier for producing convincing fake content, creating a perfect storm of misinformation during an active military conflict.

### X's Response Falls Short

In response to the surge of AI-generated conflict footage, X announced it would temporarily demonetize blue-check accounts that post unlabeled AI videos depicting armed conflict. However, the platform has not disclosed how many accounts have actually been penalized under this policy. Notably, several Iranian officials had been paying for X's premium subscription — granting them blue check marks, increased visibility, and monetization eligibility.

## Traditional Disinformation Persists Alongside AI Fakes

Beyond AI-generated content, conventional disinformation has also continued to spread. A particularly troubling case involves the attack on a primary school in Minab, Iran, on February 28, which killed over 168 people — 110 of them children. Pro-Trump accounts have repurposed unrelated conflict footage to falsely suggest the Iranian government struck its own school. In reality, footage verified by The New York Times shows a Tomahawk cruise missile — a weapon used exclusively by the United States in this conflict — hitting a naval base adjacent to the school.

### Meta Also Under Fire for AI Labeling Failures

X is not alone in facing scrutiny. On Tuesday, Meta's Oversight Board criticized the company's approach to labeling AI-generated content, stating that its current framework is "neither robust nor comprehensive enough to handle the scale and speed of AI-generated misinformation, particularly during crises and conflicts." Meta acknowledged the board's findings in an online statement.

## Detection Tools Are Struggling to Keep Up

"As AI-generated images and videos are increasingly sophisticated, users might not put into question visuals that are pushed as evidence to support pro-Iran claims when they look so real," said Isis Blachez, an analyst with media watchdog NewsGuard. "AI detection tools are not consistently successful at recognizing AI content," she added, highlighting a critical gap in the current infrastructure for combating digital disinformation.

As AI technology continues to advance faster than regulatory frameworks can adapt, experts warn that the line between verifiable reality and manufactured fiction will only grow thinner — with potentially dangerous consequences for public understanding of global conflicts.

AI-generated images and videos are overwhelming X during the Iran conflict, with millions viewing fake content as Grok spreads false information unchecked.

AI Disinformation Floods X During Iran War Crisis

large_Fake-AI-Content-About-Iran-War-All-Over-X-Politics-2191851142.jpg

medium_Fake-AI-Content-About-Iran-War-All-Over-X-Politics-2191851142.jpg

small_Fake-AI-Content-About-Iran-War-All-Over-X-Politics-2191851142.jpg

thumbnail_Fake-AI-Content-About-Iran-War-All-Over-X-Politics-2191851142.jpg

Fake-AI-Content-About-Iran-War-All-Over-X-Politics-2191851142.jpg

How AI-Generated Disinformation Is Overwhelming X During the Iran Conflict

## Google Bakes Gemini Deeper Into Workspace — And the Results Are Mixed

A major shift is underway in productivity software. As 2026 progresses, tech giants are embedding generative AI directly into the tools millions of people use every day — and Google is no exception. Despite widespread skepticism in the United States toward AI-generated content, Google is pushing forward, rolling out a suite of new Gemini-powered features across its Workspace apps. For now, these capabilities are available exclusively to English-speaking subscribers on Google's AI Pro and Ultra plans.

## What's Actually New in Google Workspace

The headline addition to Google Docs is a feature called **"Help Me Create."** Unlike its predecessor — the more modest "Help Me Write" found in the Chrome browser — this tool attempts to generate complete first drafts of documents from a single prompt. To do so, it draws on your Gmail inbox, files stored in Drive, and live data pulled from the internet.

The ambition here is significant. Google isn't just offering a writing assistant; it's building toward a future where AI handles the initial heavy lifting of putting thoughts on paper.

Google Sheets and Slides received similar upgrades, both now capable of assembling full draft presentations and spreadsheets using web data and your personal files. Google Docs also gained a structural mimicry feature, which lets users replicate the format and layout of previous documents when starting something new. Meanwhile, Google Drive received AI-generated file summaries and improved natural language search capabilities.

## Putting 'Help Me Create' Through Its Paces

Testing focused primarily on Google Docs, starting with something lighthearted: a St. Patrick's Day itinerary. Within seconds, Gemini scanned Gmail and the web to produce a short, organized plan. It even pulled up existing flight reservations to determine the correct city for the date — a detail that was equal parts impressive and unsettling. The itinerary also included suggestions for local Irish pubs. As an initial trial, the output was fast and functionally solid.

### The Real Test: Drafting a Tech News Story

The more challenging experiment involved asking Gemini to write a 600-word, first-person hands-on article about Google's own Workspace launch — the kind of piece a working journalist might produce. Press materials from Google were attached as source documents, and the request included a directive for personal insights that would help readers understand the update.

Of course, Gemini couldn't actually go hands-on with the product. What it produced instead was technically coherent but deeply generic.

*"With the latest updates to Google Workspace, we are seeing Gemini move beyond a side-panel novelty into the role of a true collaborative partner,"* the AI draft read. *"The real power isn't just in 'AI writing'; it's in the deep integration across your personal and professional data silos."*

The prose was clean, structured — and utterly lifeless. It read like a press release written by a committee, scrubbed of any real perspective or personality.

## Where These Tools Shine — and Where They Fall Flat

After generating multiple drafts across different use cases, a clear pattern emerged. Gemini's new Workspace features perform best in structured, low-stakes environments — think internal memos, marketing copy, or corporate communications where the goal is clarity over creativity. For anyone needing to produce polished boilerplate quickly, these tools genuinely deliver.

Personal expression, however, is a different story. Even after uploading samples of original writing and instructing Gemini to match that tone and voice, the results still felt generic. The AI gravitated toward safe, predictable sentence structures regardless of the stylistic input provided.

A follow-up request — asking Gemini to rewrite the draft in the voice of a seasoned tech journalist — produced a marginally improved version, but one that still leaned on the kind of buzzword-heavy framing that reads as artificial almost immediately.

*"The 'Help Me Create' engine has moved past generating generic corporate-speak, instead synthesizing live data points across your Drive, Gmail, and Chat history,"* the revised draft claimed. Given the evidence, that assertion is debatable at best.

## The Bigger Picture

There's a legitimate concern embedded in all of this. As AI writing tools grow more capable and more integrated, pressure may mount — particularly in fast-moving digital media environments — to rely on them for routine content production. Early-career writers working under high output demands could find these tools positioned less as aids and more as replacements.

For now, though, the gap between AI-generated prose and authentic human writing remains wide enough to matter. Gemini's Workspace tools are undeniably more capable than earlier iterations, with impressive speed and accurate recall from personal data sources. But they still produce writing that carries an unmistakable flatness — a by-the-numbers quality that no amount of prompt engineering has fully resolved.

The tools are useful. They are not, at least not yet, writers.

Google's Gemini-powered 'Help Me Create' drafts full documents in seconds. Here's what it does well, where it struggles, and what it means for writers.

Google Gemini 'Help Me Create' in Docs: Hands-On Review

large_gear_geminitools_GettyImages-2169339854.jpg

medium_gear_geminitools_GettyImages-2169339854.jpg

small_gear_geminitools_GettyImages-2169339854.jpg

thumbnail_gear_geminitools_GettyImages-2169339854.jpg

gear_geminitools_GettyImages-2169339854.jpg

Google's 'Help Me Create' in Docs Is Powered by Gemini — But Can It Actually Write?

## OpenAI Acquires Promptfoo to Fortify Its AI Agent Platform

OpenAI has announced the acquisition of Promptfoo, an AI security startup established in 2024 with a mission to defend large language models (LLMs) from malicious online threats. The announcement came on Monday, signaling the company's deepening commitment to building safer, enterprise-grade AI systems.

### Technology Integration Into OpenAI Frontier

According to an official blog post from OpenAI, Promptfoo's security capabilities will be woven into OpenAI Frontier — the company's enterprise platform designed to power autonomous AI agents — once the acquisition is formally completed. This integration marks a strategic step toward making agentic AI tools not only more capable, but significantly more secure.

## Why AI Security Is Now a Business Priority

The rapid rise of autonomous AI agents — systems capable of independently executing complex digital tasks — has sparked widespread enthusiasm over potential productivity gains across industries. However, this same advancement has opened new attack surfaces for bad actors seeking to exploit sensitive data or manipulate automated workflows.

This acquisition reflects a broader trend among frontier AI laboratories racing to demonstrate that their technologies are trustworthy and robust enough for deployment in high-stakes business environments.

### About Promptfoo and Its Founders

Promptfoo was co-founded by Ian Webster and Michael D'Angelo, who built the startup around tools designed to help organizations identify and address security vulnerabilities within LLMs. Their offerings include both an open source interface and a developer library, which have reportedly been adopted by more than 25% of Fortune 500 companies — a testament to the platform's real-world impact.

Despite its impressive reach, Promptfoo raised a relatively modest $23 million in total funding. The startup was valued at $86 million following its most recent funding round in July 2025, according to data from PitchBook. OpenAI has not disclosed the financial terms of the acquisition.

## What Promptfoo's Technology Brings to OpenAI

OpenAI outlined several key capabilities that Promptfoo's technology will bring to its agent platform:

- **Automated red-teaming** to proactively identify weaknesses in AI systems
- **Security evaluation of agentic workflows** to detect potential vulnerabilities before they can be exploited
- **Real-time activity monitoring** to address risk management and regulatory compliance needs

OpenAI also confirmed its intention to continue developing and expanding Promptfoo's open source offerings, ensuring the broader developer community continues to benefit from the platform's security tools.

## A Signal of the Industry's Security-First Shift

This move underscores a pivotal shift in how leading AI companies are approaching deployment. As autonomous agents become more deeply embedded in critical business operations, security is no longer an afterthought — it is a foundational requirement. By bringing Promptfoo's expertise in-house, OpenAI is positioning itself to lead not just in AI capability, but in AI trustworthiness.

OpenAI acquires AI security startup Promptfoo to integrate advanced red-teaming and vulnerability testing into its enterprise AI agent platform, OpenAI Frontier

OpenAI Acquires Promptfoo to Secure AI Agent Platform

large_GettyImages-2214107176.jpg

medium_GettyImages-2214107176.jpg

small_GettyImages-2214107176.jpg

thumbnail_GettyImages-2214107176.jpg

GettyImages-2214107176.jpg

OpenAI Strengthens AI Security by Acquiring Startup Promptfoo

## Grammarly's New AI Feature Borrows Famous Names — Without Permission

Grammarly recently rolled out a feature called **Expert Review**, promising users writing feedback inspired by some of the world's most respected authors, intellectuals, and journalists. Sounds impressive — until you dig a little deeper.

Launched in August 2025 as part of a broader wave of AI-powered upgrades, Expert Review appears as a sidebar tool within Grammarly's writing assistant. It offers revision suggestions framed as coming "from the perspective" of notable subject matter experts, giving the impression that a real authority is weighing in on your prose.

## Famous Names, Zero Involvement

As outlets like *Wired* and *The Verge* have pointed out, Grammarly presents this feedback as though it originates from well-known writers — living or deceased. Even more surprisingly, the feature reportedly invokes the names of active tech journalists working at major publications including *The Verge*, *Wired*, *Bloomberg*, and *The New York Times*.

None of these individuals appear to have been consulted, compensated, or asked for permission. When *The Verge* reached out for comment, Alex Gay, Vice President of Product and Corporate Marketing at Grammarly's parent company Superhuman, explained that these figures are referenced simply because their published work is "publicly available and widely cited."

Grammarly's own user documentation attempts to address the obvious concern, stating: *"References to experts in Expert Review are for informational purposes only and do not indicate any affiliation with Grammarly or endorsement by those individuals or entities."*

## What Makes a Review 'Expert' If No Experts Are Involved?

That disclaimer might technically cover Grammarly's bases, but it opens up a much bigger question: if no real expert is actually reviewing your writing, in what meaningful sense is this an "expert review" at all?

Historian C.E. Aubin put it bluntly when speaking to *Wired*: **"These are not expert reviews, because there are no 'experts' involved in producing them."**

The feature appears to use AI models trained on publicly available writing to generate feedback that stylistically mimics how a given expert might respond — but that's a far cry from genuine human insight or editorial guidance.

## A Clever Idea With a Credibility Problem

On the surface, the concept behind Expert Review is genuinely interesting. The idea of receiving writing feedback tailored to the style and standards of a respected author or journalist has real appeal — especially for users looking to sharpen their craft.

But the execution raises serious ethical questions. Using someone's name and professional reputation to market an AI feature — without their knowledge — blurs an important line. It risks misleading users into believing they're receiving a quality of feedback that simply isn't there, while simultaneously trading on the credibility of real people who never agreed to participate.

## The Bottom Line

Grammarly's Expert Review feature is a prime example of how AI-powered tools can over-promise through clever branding. Wrapping algorithmic suggestions in the names of respected thinkers and writers may make the product feel more authoritative — but without genuine expert involvement, the "expert" label is little more than marketing gloss.

For users seeking genuine writing improvement, it's worth understanding exactly what these tools are — and what they aren't.

Grammarly's Expert Review feature uses famous names to offer AI writing tips — but none of those experts were consulted or gave permission.