Meta and Anduril Secure Role in U.S. Army Combat Goggles Project

Meta Platforms and Anduril Industries have been selected to participate in the U.S. Army’s latest effort to develop advanced mixed-reality combat goggles. The project adds another chapter to the ongoing military push to equip soldiers with head-mounted systems that combine augmented and virtual reality.

The Army’s new program will build on the earlier Integrated Visual Augmentation System (IVAS), a multibillion-dollar initiative first led by Microsoft. That effort, despite heavy investment, faced criticism from within the military and was eventually transferred to Anduril. Since taking over, Anduril has restructured the initiative under the name Soldier Borne Mission Command, drawing on years of test data and soldier feedback.

According to available information, the new phase will use more than 260,000 hours of input gathered during the IVAS program. The Army has already committed over $1.3 billion to research, prototypes, and testing in this area, making the current contracts part of a long-running strategy rather than an isolated deal.

The collaboration also marks a renewed partnership between Meta chief executive Mark Zuckerberg and Oculus founder Palmer Luckey, now running Anduril. The two had previously pursued consumer-focused virtual reality projects but are now applying their experience to defense contracts. Bloomberg reported that Meta was included as a partner in Anduril’s proposal, although the company has not disclosed financial details of its share.

Rivet Industries, a defense technology firm led by a former Palantir executive, was also awarded a contract. The company said its deal is valued at around $195 million, suggesting that multiple prototypes will advance in parallel before the Army settles on a final system.

For the Army, the goal remains unchanged that is to field reliable mixed-reality equipment that enhances soldier awareness and mission planning. For the technology firms involved, the contracts represent a shift from consumer markets to defense applications, where timelines are long and budgets remain substantial.

Meta’s Record on Content and Advertising

While Meta’s role in the combat goggles project signals a deeper move into defense technology, the company’s broader record on ethical and moral decision-making remains under scrutiny. Over the past several years, Meta has faced repeated criticism for how it moderates content and manages political advertising, particularly during armed conflicts.

Human Rights Watch and Access Now documented widespread suppression of pro-Palestinian voices during the Gaza genocide by Israel, including removals and account suspensions affecting journalists and activists. In 2024, nearly 200 Meta employees signed an open letter criticizing what they described as systemic censorship of Palestinian content. A lawsuit filed by former Meta engineer Ferras Hamad further alleged discrimination after he raised concerns about mislabeling Gaza-related posts.

The company has also been accused of allowing ads that sought donations for Israeli military campaigns, including equipment such as drones. Reports from The Guardian, Euronews, and Al Jazeera found that dozens of ads linked to the Israeli Defense Forces remained active on Meta platforms, even when they appeared to breach internal advertising rules.

Taken together, these controversies have reinforced perceptions of inconsistency in Meta’s enforcement of its own policies, raising questions about how its role in defense projects will be viewed in light of its track record in global conflicts.

Notes: This post was edited/created using GenAI tools. Image: DIW-Aigen.

Read next: Former WhatsApp Security Head Sues Meta Over Employee Access to Sensitive User Information
by Irfan Ahmad via Digital Information World

Google’s Legal Filing Sparks Confusion Over the Open Web

Google’s lawyers told a federal court that the open web is already in “rapid decline.” The statement appeared in a filing on September 5, where the company opposed a Department of Justice proposal to break up parts of its ad business. Lawyers said forcing Google to sell its AdX marketplace would hurt publishers who depend on display advertising.

The claim stands against recent comments from Google’s leadership. In May, chief executive Sundar Pichai said the number of indexable pages had grown by 45 percent since 2023, pointing to a surge in online content. Nick Fox, vice president of Search, described the web as “thriving.” Elizabeth Reid, who leads the Search division, said last month that traffic to websites has stayed relatively stable, even with new AI features in Google Search.

Google clarifies its position

After criticism, Google explained that its filing referred specifically to advertising. Dan Taylor, vice president for global ads, said the phrase had been taken out of context and pointed to budget shifts toward connected television, retail media, and other formats. A spokesperson added that the full passage was about open-web display ads, not the entire web.

Advertising shift

Industry data shows that display ads on independent sites have lost ground for several years. Google’s own figures indicate these ads accounted for 40 percent of impressions in 2019 but only 11 percent in early 2025. Marketers now spend more on apps, video, and social platforms. For publishers that still rely on display ads, this shift has meant slower revenue growth.

Antitrust pressure

The statement comes as Google defends itself against remedies in a major antitrust case. Judges earlier found the company had tied ad services in ways that disadvantaged rivals and favored its own marketplace. Regulators are pushing for structural changes, and Google is trying to show that breaking up its ad business would harm publishers instead of restoring competition.

Reaction from industry

Observers have pointed out the conflict between Google’s upbeat public messages and the language it used in court. Some industry voices say the company is presenting different stories depending on the audience. For publishers, the core concern remains whether search traffic and advertising revenue can sustain the business models that keep much of the open web alive.

Notes: This post was edited/created using GenAI tools. Image: DIW-Aigen.

Read next: Most Content Marketers Now Use AI in 2025, With Editing Leading the Way
by Irfan Ahmad via Digital Information World

Most Content Marketers Now Use AI in 2025, With Editing Leading the Way

Artificial intelligence has moved from experiment to routine in content marketing. A new survey, conducted by Orbitmedia, shows that 95 percent of marketers now use AI tools, up from 65 percent in 2023 and 80 percent in 2024. Only one in twenty has yet to adopt the technology, making its spread faster than past shifts such as smartphones or social media.

Editing as the Primary Use

The role of AI has changed. Earlier surveys placed idea generation at the top. By 2025, editing ranked first, with about two-thirds of marketers using AI to refine drafts. Writing assistants and automated editing features have become standard in content workflows.

Few marketers turn to AI for full article writing. Only 10 percent said they use it to draft entire pieces. That share has doubled each year but remains low. Results also show that teams relying on AI for full drafts report weaker performance than those applying it for ideas or edits.

Faster Production Times

Efficiency gains are clear. Average writing time fell from 4 hours and 10 minutes in 2022 to 3 hours and 25 minutes in 2025. Among AI users, the average sits at 3 hours and 24 minutes, compared with 3 hours and 48 minutes for non-users. The 24-minute difference equals a 10 percent productivity boost.

Ongoing Concerns

Adoption has not erased doubts. Accuracy is the top concern, mentioned by 77 percent of respondents. Originality follows at 63 percent, while half raise issues with tone consistency. Ethics were cited by 41 percent, and 23 percent flagged possible SEO risks. Only 5 percent reported no concerns.

Performance and Selective Use

When asked about outcomes, 20 percent of marketers rated their content as producing strong results. Those using AI for idea generation did slightly better at 23 percent. Respondents relying on AI for full drafts reported weaker outcomes, suggesting that faster production does not guarantee higher performance.

Outlook for 2025

The survey points to a turning point. AI is now embedded in nearly all content marketing, yet its value comes from editing and support rather than replacing writers. Time savings are evident, but the strongest results remain tied to selective and strategic use.

Notes: This post was edited/created using GenAI tools.

Read next: Study Finds Human-Written Crisis Messages Viewed as More Credible Than AI
by Asim BN via Digital Information World

Study Finds Human-Written Crisis Messages Viewed as More Credible Than AI

Artificial intelligence is increasingly used in corporate communication, but new research shows it may not be suited for sensitive situations. A study in Corporate Communications: An International Journal found that crisis responses attributed to people were judged as more credible and more helpful to a company’s reputation than identical messages said to come from AI.

Testing Trust in Crisis Responses

Researchers built an experiment around a fictional company called Chunky Chocolate, which was described as facing backlash after reports that its products made customers sick. Participants read one of six possible press releases. Each message had the same content but differed in two ways: whether it was written by a person or by AI, and whether the tone was informational, sympathetic, or apologetic.

The study involved 447 students in journalism and communication programs at a Midwestern university. They evaluated the credibility of the message, the credibility of its source, and the company’s reputation after reading the release.

Human Messages Scored Higher

Results showed a clear pattern. Messages labeled as human-written were rated higher across all measures. On a seven-point scale, human sources received an average credibility score of 4.40, compared with 4.11 for AI. For message credibility, human versions averaged 4.82 while AI versions scored 4.38. Company reputation followed the same trend, with averages of 4.84 for human messages and 4.49 for AI.

Because the content of the statements was unchanged, the difference came only from how authorship was presented. Labeling a release as AI-generated lowered trust, even when the words were identical.

Tone Had Little Effect

Researchers expected an apologetic or sympathetic tone to influence perceptions. Participants did notice the different tones, but ratings of credibility and reputation did not vary much. The communicator’s identity carried more weight than the style of the message.

What It Means for Public Relations

AI already plays a role in public relations through tasks like media monitoring, content targeting, and social media management. Some suggest using it to draft press releases or respond to crises. The study points to risks in doing so, since audiences seem less likely to trust a crisis message when it is tied to AI.

Limits of the Study

The experiment used a fictional company and a student sample, which may not represent the wider public. Participants’ familiarity with digital tools and AI could also shape their views. Another factor is the explicit labeling of AI authorship, as real companies may not always disclose when AI is used.

Even with these limits, the research indicates that audiences still place greater trust in human credibility during moments of public scrutiny.

Notes: This post was edited/created using GenAI tools. Image: DIW-Aigen.

Read next:

• Artificial Sweeteners Linked to Faster Memory Decline in Midlife

• Study Finds LLM Referrals Convert At 4.87% Versus 4.6% For Search, But Scale Remains Tiny
by Irfan Ahmad via Digital Information World

Study Finds LLM Referrals Convert At 4.87% Versus 4.6% For Search, But Scale Remains Tiny

A six-month analysis of 54 websites found that traffic from large language models converts at almost the same rate as organic search. The research, carried out by Amsive, used Google Analytics 4 data from sites with validated purchases or form fills.

Conversion Rates

Organic visits converted at 4.6 percent. LLM referrals came in at 4.87 percent. On the surface that looked like a gain for LLMs, but statistical testing showed the difference was not significant. In other words, both channels brought in users who converted at nearly identical rates.

Site-Level Differences

Results varied from one website to another. Some saw LLM referrals converting better than their averages, others saw weaker performance. Just over half of the sample leaned positive for LLM, but not by a wide margin. The split highlighted how much outcomes depend on how AI tools select and surface content.

Higher-Volume Sites

Filtering for larger websites, those with at least 100,000 sessions and enough LLM traffic to test, produced a bigger gap: organic at 5.81 percent and LLM at 7.05 percent. Even then the edge failed to clear statistical tests. The analysis showed that the apparent lift could be explained by random variation.

Business Models

Breaking the data into B2B and B2C websites did not change the picture. B2B sites converted at 2.03 percent from LLM referrals and 1.68 percent from organic. B2C sites converted at 10.31 percent from LLM and 8.50 percent from organic. Neither difference was large enough to be reliable once tested for significance.

Industry Patterns

By industry, outcomes were mixed. Financial services and travel recorded higher conversion rates from LLMs. E-commerce and consumer services leaned toward organic. Because sample sizes in each vertical were small, no firm conclusion could be drawn.

Traffic Share

The study found scale to be the critical factor. LLM referrals accounted for less than one percent of total sessions across the dataset. Organic search made up nearly a third of all visits and conversions. In fact, about nine out of ten websites saw LLM traffic contribute less than 0.6 percent of sessions.

Study Limits

The research measured only macro conversions such as purchases and lead forms. It did not track how many leads became paying customers. Conversions were counted on a session basis and attributed to the last click, so earlier touchpoints were not included.

Key Takeaway

For now, organic search remains the leading channel for both scale and consistency. LLM referrals may grow in importance as usage expands, but current evidence shows they are not outperforming search in conversion terms. Businesses may want to monitor LLM traffic closely while continuing to treat search as the foundation of their digital strategy.

 Notes: This post was edited/created using GenAI tools.

Read next:

• Bad Sign-Up Flows Cost SaaS Companies Customers Before They See the Product

• How Many Prompts Can You Run on Gemini Each Day? Google Finally Sets the Numbers
by Irfan Ahmad via Digital Information World

Bad Sign-Up Flows Cost SaaS Companies Customers Before They See the Product

With SaaS, you've got no second chance to make an entrance. Customers come through the door, sign up, and demand to be given instant access. Anything less and they disappear forever. A new Frontegg survey verifies just how much you stand to lose: 15% of SaaS users never come back after a less-than-perfect login experience. Frontegg's First Login Benchmark Report surveyed 439 SaaS users to find out what they crave most from onboarding and why companies get it fundamentally wrong. What their findings reveal is just how little SaaS products need to do in order to earn user trust and, having lost it, how rarely the window of opportunity reopens.

The First Impressions Aren't Skin-Deep Quiz

Onboarding is not a handshake anymore. It's a background check, trust evaluation, and product demo all in one. Nearly half of the users (48%) reported that they had abandoned a SaaS service due to registration taking too long. They would have already decided to return or not by the time they reached the dashboard. 36% churned on confirmation by e-mail step, and 21% insisted on instant access on sign-up. This go-fast strategy isn't one taken in an attempt to compromise on security at all. It truly was the case, as speed and security did factor into 46% of those making their initial impression of a SaaS solution. Just 32% prioritized security over speed. That fast and stable combination is hard to find and nearly impossible to regain.

Why Login UX Can Make or Break Trust

Over half the users (58%) would not even subscribe to a SaaS application whose logon process did not look above board. That's a harsh term, but it's the best word to use for the gut-level feeling many users get when design, language, or interaction flow triggers a sense of unease. It isn't rational, but perceptions govern. The majority of the irritations were old favorites. CAPTCHA was the most common top-of-the-list trust-breaking authentication process, seconded by convoluted multi-step identification verifications and cumbersome email verifications. The processes may lead to added security, but are most typical indicators of lack of design maturity or cumbersome flow if done ineptly. And the psychological impact is real. Twenty-one percent of the surveyed users admitted to rage-quitting onboarding by bailing on the entire process right there out of frustration long before even getting to experience the product. Those numbers show a bigger pattern. Login UX was once ambient. It then turned into the product experience itself.

The Long-Term Relevance of Login Interfaces

The most secure login features, most strongly linked in users' minds with safety and professionalism, come as no surprise yet demonstrate a persistent gap between desired and delivered.

Two-factor authentication (2FA): 74% of users reported feeling safer when it's enabled

Email confirmation: 23% consider it a sign of trust if executed well

Single sign-on and social login: 20% and 11%, respectively, value the convenience of them

These statistics indicate why intelligent design and simplicity are key when it comes to establishing user confidence. Users do not want the verification to be omitted completely, but rather quick, secure, and transparent.

Cost of a Bad Experience per Person

SaaS products have only one chance. 53% of the users would return to a SaaS product after a bad login experience, according to Frontegg, yet 15% would never return. Even if they do remain, they use it fractionally. Within a 30-day trial, over half (51%) had used 25 to 50% of the product features. That extent of selective adoption does more than limit the initial impression. It holds product stickiness and long-term satisfaction at bay. It is not just first-week churn. The negative sign-up impact extends into usage, word of mouth, and upsells.

What SaaS Teams Can Do Differently and Learn from

Understanding where users drop off is a crucial first step, but the tougher part is determining how to remove the friction causing the drop off, without implementing something that adds friction.

Here are some areas of focus for SaaS teams creating a login flow or reimagining one:

Removing friction without decreasing security. A fast way to access your app is important, but fast access will be no good and will not build trust if it amounts to a sketchy login. Users are willing to accept 2FA if it works properly; ineffective versions of Captcha or multi-step verification will likely drive users in the opposite direction.

Making verification frictionless. Email verification is a common drop point; real-time verification messages or progressive disclosure might alleviate friction at this delicate moment of onboarding for your users.

Treating the login UX process like the design of the core product. All screens and steps should feel purposeful for users. Confusion over where to go next or confusion over what the progress appears as can damage trust faster than slow-loading issues.

Testing sign-up as a potential new user. Internal teams often gloss over and ignore obstacles that new users frequently encounter. Testing, along with real user feedback, will provide the insights that metrics can miss.

The Big Picture: Trust Before Features

SaaS functionality and workflow take months or years to build. It will be useless if the users never even go beyond the front door. Frontegg data exemplifies the cultural pivot. Onboarding isn't just getting users past the front door. It's the first promise of value, security, and empathy. And when virtually one in every five users rage-quits on the first day and an additional 15% ghost after the one bum login, the margin of error basically does not exist anymore. For SaaS companies, that's it. Good onboarding is not a nice-to-do. Growth. Retention. Reputation. Done. The virtual handshake is now. SaaS companies that wish to retain customers long enough to show them what they're providing must trust customers immediately and often. That starts on the login page. The full First Login Benchmark Report results reveal more about why and when customers churn SaaS products prior to use. Get the full study here to read the recommendations and benchmarks.

Read next: How Many Prompts Can You Run on Gemini Each Day? Google Finally Sets the Numbers

by Irfan Ahmad via Digital Information World

Why AI Chatbots Still Hallucinate: Researchers Trace Errors to Training Data Gaps and Misaligned Benchmarks

Artificial intelligence tools are now used in classrooms, offices, and customer support desks. Yet they carry a flaw that refuses to fade. Ask a chatbot a simple factual question, and it may deliver a confident answer that turns out to be wrong. Researchers at OpenAI, joined by collaborators at Georgia Tech, say they now have a clearer picture of why this happens.

Where the Mistakes Begin

Large language models are trained by scanning enormous volumes of text and learning to predict what word should come next. That process gives them fluency, but it also builds in errors. The team’s paper explains that even with perfectly clean training data, mistakes are mathematically inevitable.

Some facts are simply too rare for a system to learn. A birthday that appears once in a dataset, for example, provides no pattern the model can generalize. The authors call this the “singleton rate.” High singleton rates mean a model will almost certainly invent details when asked about them. This is why common knowledge tends to be correct, while obscure details often come back scrambled.

From Exams to Algorithms

The training phase is only half the story. After that, models are fine-tuned to better match human expectations. But the way they are tested keeps the cycle going.

Benchmarks usually grade answers as right or wrong. There’s no credit for admitting uncertainty. A chatbot that says “I don’t know” is punished as harshly as one that blurts out something false. Under that system, guessing is the smarter move. Over time, models are effectively trained to bluff.

The researchers compare this to multiple-choice exams. Students who leave blanks score lower than those who make lucky guesses. AI models, shaped by similar scoring, act in much the same way.

When Models Go Wrong

Examples from the study illustrate how deep the problem runs. One widely used model was asked for Adam Kalai’s birthday — Kalai being one of the paper’s authors. It gave three different dates across separate attempts. None were right, and it had been told to answer only if certain.

In another test, a system failed at counting the letters in a word, producing results that made little sense. These cases show both the arbitrary fact problem and what the authors call poor model representation, where the structure of the system limits its ability to handle simple tasks.

Changing the Scoreboards

The researchers argue the solution lies in evaluation. Instead of rewarding risky guesses, new benchmarks should penalize confident wrong answers more than admissions of uncertainty. One option is to grant partial credit when a model holds back. Another is to set confidence thresholds in the test instructions, telling the model to answer only if it reaches a defined level of certainty.

This echoes older exam systems where wrong guesses were penalized, discouraging blind attempts. The same principle could shift AI development toward models that value accuracy over bravado.

Limits and Outlook

The study makes clear that hallucinations will not vanish completely. Some questions are inherently unanswerable because the data is missing, ambiguous, or too complex. But better testing could reduce the most damaging errors and build greater trust in AI systems.

The broader point is that hallucinations are not random glitches. They are the product of how models are trained, and more importantly, how they are judged. If the industry changes the scoreboards, the behavior of the models is likely to follow.

Notes: This post was edited/created using GenAI tools. 

Read next:

• AI Models Can Now Run Ransomware Attacks on Their Own, Study Finds

• Secure Online Transactions and Business Models in E-commerce and Marketplaces

• Chatbots Are Spreading More False Claims, NewsGuard Report Shows

by Irfan Ahmad via Digital Information World