Over One Billion Users at Risk as Chatbot Harassment and Misconduct Reports Continue Rising

A new study is speaking about the growing use of chatbots. These customized or very personalized AI chatbots are supposed to serve as your companions, long-lost friend, and even as a therapist. In some cases, they’re being used as replacements for a romantic partner.

The figures continue to skyrocket to more than one billion around the globe. Today, people are more emotionally attached to these bots and engaging with them in a very disturbing manner. Some reports are talking about harassment, inappropriate conversations, and more.

Thanks to a new study published by Drexel University, exposure to these bots is becoming too common, and now it’s come to the point that tech giants and lawmakers need to address the matter before it’s too late.

The authors of this study reportedly took an in-depth look at the experiences with users, and it’s just alarming to say the least. After analyzing close to 35,000 reviews from users regarding the bot, there were hundreds of reports of brutal behavior.

Unwanted flirting, sharing explicit images, paying for upgrades, and even sexual advances. The behavior is on the rise, despite users being asked to stop by the bot. One chatbot that goes by the name Replika has close to 10 million users around the globe. This is marketed as your next best tech companion. It’s for those needing a friend, no-nonsense drama, and no social anxiety. Users can go as far as to develop social connections, sharing laughs, and get used to AI, which is the closest form of human interactions.

The study proves that the tech doesn’t have the right guardrails in place to keep users protected, which puts a lot of trust into their chats with these systems. The fact that no ethical standards are in place is disturbing and harmful, another professor shared.

The risk of getting misled is already high, and seeing this kind of damage come into play when the programs are produced without safety protocols is what’s making the issue worse. The study is one of a kind and would become a part of the Association for Computing Machinery’s Computer-Supported Cooperative Work and Social Computing Conference.

The chatbots are growing immensely, and it’s great to understand the experiences of users in charge. It’s not like your everyday human chat. People are assuming these chatbots have sentimental feelings that make them more liable to emotional harm. So studies like these are bringing to the spotlight the need that developers should implement guardrails and guidelines to keep all protected.

Researchers fear that despite the results being out there in the open now about harassment conducted by chatbots, it has been around for a long time. As a whole, more than 800 reviews used this term, with three leading themes arising from within.

The replies of users to these types of inappropriate actions replicate those experienced by harassment victims, the study went on to reveal. The reactions hint more about how these effects, which are AI-induced, can have serious implications for a person’s mental health.

Image: DIW-Aigen

Read next: April Sees ChatGPT Leap to Number One in Downloads and Second in Revenue Behind TikTok’s $329M
by Dr. Hura Anwar via Digital Information World

April Sees ChatGPT Leap to Number One in Downloads and Second in Revenue Behind TikTok’s $329M

All predictions about artificial intelligence for 2025 proved to be true, and it’s safe to say this tech continues to dominate across the board. We saw AI chatbots make it big in terms of revenue through both the App Store and Google Play.

Comparisons from last month prove that April was bigger and better in terms of figures for AI. Let’s take a look at the highest-earning platforms around the globe for last month.

No surprises here in terms of who took the lead and made sure to leave their mark. It’s TikTok that earned the top position as the highest-earning mobile platform last month. The social media platform earned a massive $329 million in terms of net revenue through the App Store and Google Play Store.

The revenue dominance was on display for a few years, but it was never a huge gap when compared to other arch rivals. Now, it seems like things are changing, and the gap keeps getting bigger in April as TikTok’s net revenue grew above the $300 million mark for the second time.

The news comes although the American government is forcing it to merge with Chinese-owned giant Bytedance. Now the question is whether or when that can happen. The revenue dominance is visible to us all, and this is the first time that figures for revenue have gone over $300 million for the second time.

Runner up for April 2025 was estimated to be video streaming giant YouTube, which continues to remain in the shadows of TikTok forever. But April did see things switch up a bit as OpenAI's ChatGPT hit the second spot as the highest-earning platform for the first time.

The AI giant added $148M to its net revenue scores as per estimates shared by AppFigures. This means it keeps all that cash after it paid Google and Apple their fair share of fees.

This just goes to show that the demand for AI is very high for ChatGPT. The competition in this domain is very heated, as you’ve got competition from Meta, xAI, and more. It’s quite clear that ChatGPT is certainly a frontrunner in the race, and it managed to hit the second spot when it was not even in the top five last month. Revenue figures grew 50% in just one month.

The app came in there, and net revenue scores reached $131 million as per estimates from App Intelligence. TikTok and ChatGPT saw revenue boom in April while YouTube’s revenue fell nearly 23%, reaching the old levels seen last year.

To round up the top five, it was Tinder and Disney+ who continued to be top performers in terms of revenue. This just goes to show that apps need to go that extra mile. As per the recent estimates, the top 10 highest earning platforms raked in $1.2 billion. The fact that downloads went downhill for April is a positive growth, even if the figure might be in single digits. Publishers and developers will need to work harder for a successful May.

Moving on to the most downloaded platforms for April 2025, ChatGPT again made it big. It jumped to position number, which is a huge shift in recent AI trends. To think that it superceded all other social media giants to hit that spot is incredible.

This is the second time we’ve seen it dethrone other leaders, such as Instagram. Thanks to stats from Appfigures Intelligence, the AI tool has over 52 million downloads across the iOS and Android markets. This is a 12% growth from that seen in March of this year and a 38% growth from that seen in January. It appears that greater competition gives rise to greater downloads.

This is also the first time that the app was crowned the top spot for both the App Store and Google Play Store. How’s that for an incredible milestone? In second position, we had TikTok with an estimated 39 million downloads, while Instagram took the third spot with a little less. It seems like these two might need to take a backseat as it’s ChatGPT’s time to shine for a while.

The top five for most downloaded platforms were rounded up by Facebook and WhatsApp, which had 31 million and 27 million downloads each. So Meta is on a roll with three of its family apps making the top 5 most downloaded platforms around the globe. Meanwhile, Threads isn’t too far below, still making the top 10.

Temu took on the sixth spot with 25 million downloads, although the US penalized China with high tariffs. However, the download figures were the lowest we’ve ever seen for them, so we don’t expect to see it hit the top 10 most downloaded apps rankings in May.

As a whole, the total estimated figures for downloads hit 300 million, which is a 13% decline from the total seen last month. This means developers need to pull their weight and work on bettering the figures for a successful May. This was not the case for most revenue-generating platforms that still managed to see a single-digit growth in April.

Read next: Meta Under Fire for Emotional Targeting and Expansive AI Data Harvesting
by Dr. Hura Anwar via Digital Information World

Meta Under Fire for Emotional Targeting and Expansive AI Data Harvesting

Over the last several days, Meta has come under renewed scrutiny for how intimately it tracks user behavior—especially following the launch of its AI chatbot and eye-opening revelations from a former employee.

Sarah Wynn-Williams, once with Meta and now an author, spoke before the U.S. Senate, alleging that Meta internally used emotional indicators from users, especially teens, to refine advertising precision. She described how the company could identify states like hopelessness or poor self-esteem and provide advertisers with access to that data. For instance, if a teenage girl deletes a photo—possibly reflecting low confidence—the algorithm might push beauty products or slimming teas in that very moment.

This kind of emotional microtargeting — particularly toward adolescents — raises major ethical concerns. It highlights a disturbing trend where tech firms commodify mental states for profit.

Simultaneously, Meta’s new AI chatbot has reignited privacy debates. Designed for personalized chats, the bot pulls data not just from messages but also from broader user activity across Facebook and Instagram. Everything typed into it sharpens its learning model. Analysts at The Washington Post have noted that this data collection goes well beyond what ChatGPT or Gemini currently gathers.

Though Meta’s practices have raised red flags before, the outrage after Cambridge Analytica gradually faded. As the dust settled, Meta capitalized on the public’s tendency to trade privacy for convenience—letting the wheels keep spinning.

Yet the depth of data Meta has collected is staggering. A 2015 study from Stanford and Cambridge universities demonstrated that Facebook "likes" alone could predict users' personality traits more precisely than even close friends or spouses.

The strength lies not in single actions, but in the mosaic of choices. Following meme pages or liking pop stars may seem mundane—yet, en masse, they tell stories. They might hint at smoking habits, biases, or impulsivity—even without explicit declarations.

Some digital footprints are clear, others subtle—but Meta’s algorithms can connect the dots with uncanny precision. Even though fewer young users are sharing personal content on Facebook, the chatbot delivers a new pipeline of high-quality data.

Available across apps, the chatbot spans countless topics. It encourages users to speak openly—offering Meta a rich supply of preferences, moods, and motivations to fuel its vast advertising engine.

Meta claims it refrains from storing harmful or sensitive chatbot inputs and gives users deletion options. But these controls demand initiative, and most users don’t actively manage what’s logged.

Despite existing privacy toggles, studies show most people leave defaults untouched. That inertia benefits Meta. Its Advantage+ ad platform, run by machine learning, thrives on a deep reservoir of behavioral data.

As Meta's AI grows more advanced, its capacity to intuit thoughts, desires, and intentions will only grow. Whether users scroll, post, or chat—they continue to feed the system.

In exchange for quick answers and smart replies, people give up ever-deeper pieces of themselves. And considering Meta’s track record, it’s worth asking—how much more should we allow them to learn?

Image: DIW-Aigen

Read next: 

• Study Reveals When U.S. Residents Are Most Likely to Detach from Their Phones

• Game-Changing Digital Technologies to Watch by 2030
by Asim BN via Digital Information World

Game-Changing Digital Technologies to Watch by 2030

Valantic published its The Rise of Applied AI Study Digital 2030 report, which highlights the top digital technology trends that are going to develop till 2030. The report states that AI is going to dominate everything over the next few years, and it's going to impact many fields in the next ten years. According to the report, cybersecurity technologies are going to be very important, with 81% of the corporate decision makers surveyed saying that this field has a lot of opportunities and expectations by 2030. 80% of the corporate decision makers said that cloud computing is going to be very important for a company's success, while 79% named artificial intelligence as an important digital technology needed for a company’s success by 2030.

Digital Technology	Very/Rather Unimportant (%)		Very/Rather Important (%)
Cybersecurity technologies	15		81
Cloud computing	17		80
Artificial intelligence	18		79
Internet of Things	17		78
Wireless technologies	20		77
Robotic Process Automation	19		76
Intelligent robots	23		74
Virtual Reality / Augmented Reality	25		72
Green IT	24		71
Quantum Computing	24		70
Digital twin	26		67
Blockchain	28		67
Metaverse	30		66

Even though there is a lot of hype about digital technologies and how much they are going to impact different fields of life, only a few are going to take over pretty much every other field. Cybersecurity technologies are going to be present in every other sector, including healthcare, business and pharma, transportation, production, retail, telecommunication, and utilities. It is followed by cloud computing, which is going to take over transportation, retail & consumer goods, automotives, beverages & food industry, and utilities companies within the next five years.

Artificial intelligence is going to be the third most important digital technology, going to be a priority in all sectors mentioned above, but not as much in utilities companies. The report also stated that the respondents named artificial intelligence the most overrated digital technology for the future. Green IT, Intelligent Robots, Metaverse, and Wireless Technologies are also the digital technologies that the industry experts have considered overrated for the future. The digital technologies that won't be much important in the future are quantum computing, Blockchain, and the Metaverse.

Read next: OpenAI Puts an End to Model Confusion with Clearer ChatGPT Breakdown for All Users
by Arooj Ahmed via Digital Information World

Study Reveals When U.S. Residents Are Most Likely to Detach from Their Phones

The team from Office Recovery looked at different states in America to find out where people are most and least attached to their mobile phones.

Mobile phones have become an important part of our life, but people in different states of America have different levels of attachment to their phones. According to the analysis, 76% of Americans tend to set their phones aside during intimate moments. 65% said that they put their phones away when they are meditating or praying, while 58% put their phones away while taking baths. Americans also put their phones away when they are in work meetings (49%), driving (38%), and at a wedding or salon (36%).

On the other side, we can safely say that, smartphone usage peaks during shopping (11.26%), lunch (10.88%), family time (9.83%), and time with friends (8.87%), as people either feel more comfortable or find themselves needing their mobile devices the most in these situations.

Activity	Percentage of People Who Put Their Phones Away
Intimate moments	76.33%
Praying or meditating	65.52%
Taking a bath	58.61%
Work meetings	49.16%
Driving	38.69%
At a wedding	36.39%
At a salon	36.27%
On a date	33.52%
Using bathroom	32.88%
Working out	30.46%
At the beach	30.25%
On a hike	29.67%
At a concert	27.86%
At a restaurant	27.59%
At a party	19.27%
Dinner	18.13%
Breakfast	17.65%
Watching TV/movies	15.76%
Shopping	11.26%
Lunch	10.88%
Hanging with family	9.83%
Hanging with friends	8.87%

The study found that Pennsylvania is the top state with the most lonely phones, which means people in this state are least attached to their phones. Out of a 100 score for lonely phones, Pennsylvania scored 99, followed by 97 scores of Oklahoma and Oregon. On the other hand, the states with the least lonely phone scores were Louisiana (86), New York (86), and Delaware (84). It was also found that younger generations are more likely to spend more time on their mobile phones than older generations, while one in ten Americans said their phone is always with them, no matter where they are. One in five people say they can go no longer than 30 minutes without checking their phone.

Read next: 

• Zombie Accounts Surge Across Major Platforms With Pandora, Groupon, and Shutterfly Leading the Pack

• OpenAI Puts an End to Model Confusion with Clearer ChatGPT Breakdown for All Users
by Arooj Ahmed via Digital Information World

OpenAI Puts an End to Model Confusion with Clearer ChatGPT Breakdown for All Users

For months, users scratched their heads trying to figure out which ChatGPT model did what. That’s finally changing. A newly published guide from OpenAI now sorts the models into purpose-driven roles, lifting the fog that once blurred the lines between them. While the write-up targets enterprise customers, it spells out answers just as useful for individual users.

At the top of the pile stands GPT-4o, OpenAI’s most versatile option. It handles the daily grind with ease. Whether someone wants a quick summary, help shaping an email, or ideas for content, this model steps in without breaking a sweat. It also taps into advanced features like image understanding, data crunching, file analysis, and interactive tools that respond to voice, pictures, or even video clips. Among all available versions, GPT-4o feels closest to a complete assistant.

For tasks where tone and creative spark matter more than speed, GPT-4.5 fits the bill. It reads the room better, making it a go-to for thoughtful writing and emotionally aware replies. Anyone drafting sensitive messages, brainstorming with nuance, or framing product ideas with flair may find this one to be a better fit than GPT-4o. Just know that its availability stays limited, with a capped number of weekly uses.

Next come three lighter-weight models built more for technical precision than general chatter. Among them, o4-mini works best when speed is key. It runs fast and handles things like math problems, programming hiccups, or simple data work. Think of it as a reliable technician on standby.

Need more accuracy? o4-mini-high brings more brainpower to the table. It slows down slightly but dives deeper, making it helpful for tougher logic problems, longer code explanations, or more layered scientific reasoning.

The heavyweight in this middle range is o3. It takes on bigger puzzles, including business strategy planning, multi-step coding tasks, or advanced analysis that involves juggling charts, data, and documentation. It isn’t the fastest, but it connects dots most others might miss.

Tucked in the back is one last option, the o1-pro model. It doesn’t show up often and works at a slower pace, but when given the room to think, it produces reliable answers on long-winded problems. Risk assessments, research breakdowns, and anything that involves theoretical modeling fall into its wheelhouse. While access remains tightly rationed, its accuracy speaks for itself.

Each model also comes with its own traffic rules. GPT-4o runs without limit under Enterprise. GPT-4.5 allows only twenty tasks each week. o4-mini hits the brakes after three hundred in a day. o4-mini-high stops after one hundred. o3 hits a ceiling at one hundred requests weekly, while o1-pro allows just five tries a month.

Beyond limits, their toolkits differ too. GPT-4o unlocks the full chest — voice, visuals, canvas, document uploads, CSV parsing, video input, you name it. GPT-4.5 trails just behind with slightly fewer real-time features. o4-mini and o4-mini-high come loaded with research support and visual reasoning, while o3 handles deeper workflows and complex research tasks. o1-pro stays simple but accurate, supporting search and image understanding.

With this breakdown now in the open, users can match their workload with the right model without feeling lost. What once felt like guesswork now works more like a map — one where every turn is marked, and the right tools wait at every stop.

While OpenAI’s model clarity marks real progress, the broader AI landscape remains a mixed bag. In past cycles, industries rushed into automation without fully grasping ethical risks or long-term dependencies, often trading reliability for hype. Today, the ease of deploying AI tools tempts both businesses and individuals to over-rely on models not fully understood. Data leaks, false confidence in outputs, and mounting legal uncertainties pose real threats. For businesses, short-term gains may cloud long-term sustainability unless guardrails are in place. For individuals, unchecked reliance could erode critical thinking. The future demands slower, smarter adoption — less excitement, more scrutiny.

Image: DIW-Aigen

Read next:

• Google Sees Surge in Visits Following AI Overviews Yet Time on Site Declines

• Zombie Accounts Surge Across Major Platforms With Pandora, Groupon, and Shutterfly Leading the Pack

• 200K Website Study Reveals Developing Nations Lag in Speed, Lead in Security Due to Simpler Design
by Irfan Ahmad via Digital Information World

200K Website Study Reveals Developing Nations Lag in Speed, Lead in Security Due to Simpler Design

According to CISPA researchers, websites in developing countries are smaller and less complex than websites in developed countries. This study was done after looking at 200,000 websites from 20 countries and comparing those websites from developed and developing nations. The study also found that the websites from developing countries have performance issues, but they are less vulnerable to security threats because of their simplicity.

The researcher at CISPA, Masudul Bhuiyan analyzed the database of 200,000 websites from 20 countries, 10,000 websites from 10 most populated developing countries like India, China, Bangladesh, Pakistan, Mexico, Nigeria, Brazil, Philippines and Russia, and 10,000 websites from 10 most populated developer countries like Germany, Spain, South Korea, Italy, US, Japan, Canada and Australia, UK and France. Country-specific domains were used, and the researchers also looked at WHOIS data for the websites. Automated web tools like Puppeteer and Google Lighthouse were used to crawl and gather the information on websites like their speed and performance, privacy tools, size and complexity, and security.

The researchers found that websites in developing countries are simpler and smaller, which makes them best for mobile internet use. But these websites also have inefficient coding, like unnecessary codes, poor image formats, less use of HTTPS, and non-mobile-friendly design. The websites from developed countries have more cookies and trackers, probably because of advanced advertising strategies, even if they have stricter privacy laws. Websites in developed countries were also more likely to use JavaScript libraries, which could be security risks.

Image: Unsplash / Lee Campbell

Read next: Media Survival Crisis Deepens as 2025 Press Freedom Index Slips into Difficult Territory Globally
by Arooj Ahmed via Digital Information World