Tech

TikTok’s U.S. user base stabilizes after rocky start

TikTok’s U.S. joint venture seems to have survived a turbulent rollout with minimal change in usership, as early narratives of a mass user exodus prompted by service outages and censorship concerns now appear overstated, according to new figures.

Survey data from market intelligence firm Sensor Tower show that, despite a surge in deletions following the announcement of TikTok’s U.S. joint venture on Jan. 23, the average number of TikTok’s daily active users in the U.S. remains around 95% of its usership compared to the week of Jan. 19-25.

The joint venture — officially the TikTok USDS Joint Venture — was established in compliance with U.S. President Donald Trump’s executive order mandating the divestiture of TikTok in the U.S. from its Chinese parent company ByteDance.

While ByteDance retains a 19.9% stake in TikTok’s U.S. operations after the agreement, Oracle, Silver Lake, and Abu Dhabi-based investment firm MGX each own a 15% share, with the remaining shares divvied among several other firms.

Following the announcement, users were quick to express discontent over TikTok’s new ownership.

The deal drew scrutiny, with prominent figures like Sen. Bernie Sanders (I-Vt.) raising concerns about cronyism over Oracle co-founder and Chief Technical Officer Larry Ellison’s involvement.

Following the joint venture’s announcement that Ellison’s Oracle would “retrain, test, and update the content recommendation algorithm on U.S. user data”, online speculation mounted that TikTok would begin mining user data or promoting content supportive of Trump’s policy positions.

Such concerns spiked on Jan. 25, with users claiming that TikTok was suppressing content critical of controversial Immigration and Customs Enforcement operations, and censoring buzzwords like ‘Epstein’ on the platform.

Last month, CNBC confirmed that messages containing the word “Epstein” triggered an error message, but was unable to independently verify broader claims of political censorship.

Asked about the issues, a spokesperson for the TikTok joint venture told CNBC in January that the platform does not prohibit sharing the name ‘Epstein’ in messages and that it was investigating why some users are experiencing the problem, among others.

CNBC reached out to the White House and TikTok for comment but did not receive a response by publication.

Engagement metrics unchanged

Although TikTok attributed last month’s disruptions to power outages, the glitches “no doubt impact[ed] how and what content was being served, even without any intent or motive,” according to Jim Johnston, partner at law firm Davis+Gilbert LLP.

Yet despite various user pledges to boycott the platform over apparent political suppression, engagement metrics among U.S. users suggest there has been little sign of a mass exodus.

The average daily time spent by American users on the platform has since returned to around 80 minutes, after dipping to an average of 77 minutes during the week of the reported disruptions, according to Sensor Tower data.

Additionally, while deletions spiked after the reported disruptions, they tapered off the following week, suggesting a temporary surge rather than a sustained boycott of the app.

“It is plausible that the short-lived rise in observed uninstalls was due to an attempt to troubleshoot the app,” Abraham Yousef, senior insights analyst at Sensor Tower, told CNBC, as the number of uninstalls followed by re-installations on the same day surged more than 70% on Jan. 25 from the day before.

While Yousef grants that the data suggests a “slight impact to overall usage” in the weeks after the Joint Venture was announced, there is no clear indication of a structural shift in user trends, as many sites touted as alternatives to TikTok have also struggled to sustain interest.

According to Sensor Tower, the number of new installs for UpScrolled – a social media platform that offers an algorithm free of automated systems that filter out content from some users known as shadow banning – surged by about 770% from the previous week, with more than 955,000 new U.S. downloads over the week of Jan. 26 to Feb. 1.

New UpScrolled downloads, however, fell sharply by about 80% the following week, bringing in only around 191,000 new users. In comparison, TikTok registered 870,000 downloads over the week of Jan. 26 to Feb. 1, and around 800,000 the following week.

Similarly, new downloads of other alternative platforms such as Skylight Social and Red Note respectively declined by 96% and 33% week-on-week from the week of Jan. 26.

Tenuous evidence of mass exodus

Sensor Tower’s user data more fundamentally seems to suggest that beyond anecdotal claims, users have largely been unable to identify tangible changes in TikTok’s American operations, or at least, not enough to meaningfully shift user sentiment.

“The idea of a mass exodus from TikTok now looks overblown,” Kelsey Chickering, principal analyst from Forrester, told CNBC. “Anecdotally, most users say the app feels largely the same – the algorithm hasn’t meaningfully changed, and the experience is still strong,”

While some American users may have perceived changes in the operation of their TikTok algorithms, “some changes to content suggestions are bound to occur simply due to the changed data set,” according to Johnston, referring to the Joint Venture’s announcement to retrain the algorithm on U.S. data.

But while analysts have been unable to find evidence that TikTok’s new American owners have engineered the platform in their favor, this is not a foregone conclusion.

According to Johnston, there are at least three notable changes to TikTok’s new terms of use, including the platform’s ability to collect precise location data from enabled devices, its collection of data on interactions with artificial intelligence tools on the app and its explicit integration with ad networks.

Although there has been no hard evidence of its occurrence, it remains technically possible to adjust TikTok’s algorithm to enhance or diminish the impact of certain types of content on recommendations, Johnston said.

Chickering adds that under its new owners, TikTok has more control over what shows up on American feeds, but this control, according to Chickering, is where TikTok’s opportunity – and risk – lies.

“If moderation starts to feel politically slanted or misinformation isn’t adequately addressed, the platform could face backlash from users and advertisers alike,” Chickering said, “We’ve seen this play out before: Twitter’s shift to X is a recent reminder of how quickly trust can erode.”

For now, however, the discontent from TikTok’s American users that marred its first few weeks under new ownership seems to have largely subsided.

As Chickering notes, “we’ve seen time and time again, if the product works, users tend to stick around regardless of who owns it.”

— CNBC

Tech

PayPal Data Breach Exposed User Data for Six Months: What to Know

PayPal officially disclosed a significant data exposure incident involving its PayPal Working Capital (PPWC) application. In a newly circulating security incident, PayPal confirmed that sensitive customer information was exposed for nearly six months in 2025 due to a software flaw in one of its business financing tools.

The breach affected users of PayPal’s Working Capital loan application, exposing a wide range of personally identifiable information, including highly sensitive data such as Social Security numbers and dates of birth. According to PayPal, the incident originated from a coding error within the PayPal Working Capital (PPWC) loan platform.

The company says the issue persisted from July 1 until mid-December 2025 before being identified and rectified

PayPal data breach exposed sensitive user data for six-month period; what you need to know

What should users do?

It is recommended for users to take following steps:

Enrolling in credit monitoring services
Placing fraud alerts or credit freezes if necessary
Updating passwords across financial accounts
Being cautious of unsolicited communications

PayPal’s latest disclosure adds to a growing list of high-profile data exposure incidents in the financial sector, underscoring the risks associated with digital financial services in an increasingly digital economy. The incident highlights ongoing challenges in the wake of rising security threats. The prolonged duration of these security challenges-nearly half a year-raises questions among regulators and customers regarding detection capabilities and internal monitoring processes.

By The News Digital

Tech

Google’s new AI lets you walk and fly,

What would it feel like, to walk through a desert that you dreamed about? Or maybe, fly over the fantasy city you’ve always wanted to live in. Maybe, just maybe, you can drive through a forest, all created from a single sentence. Google DeepMind’s new experiment, Project Genie, is trying to make that possible.

Announced on January 29 via Google’s Innovation and AI blog, Project Genie is an experimental research prototype that turns simple text or image prompts into interactive, explorable virtual worlds.

What exactly is Project Genie?

Project Genie is powered by Genie 3, Google DeepMind’s latest “world model” AI. Unlike traditional image or video generators, Genie 3 creates environments that respond to your actions in real time. You can move through these worlds by walking, flying, or driving, while the AI generates what comes next on the fly.

According to Google, these worlds can range from realistic settings, like forests and deserts, to more imaginative, fantasy-style environments.

Why this is a big deal

World models like Genie 3 represent a shift in AI research. Instead of producing static images or short clips, they aim to simulate living, changing environments that maintain continuity and logic as users explore them.

Google DeepMind unveiled Genie 3 in 2025, calling it a breakthrough because it can sustain interactive worlds for several minutes, rather than just a few seconds.

What you can do inside Genie

According to Google’s blog, Project Genie includes:

World sketching: Create worlds using text prompts or images
World exploration: Move freely as the AI generates new terrain in real time
World remixing: Modify or combine worlds and download them as videos

As Google explains, “Unlike static 3D snapshots, Genie 3 generates the path ahead in real time as you move and interact with the world.”

Who can try it?

For now, Project Genie is available only to Google AI Ultra subscribers in the United States aged 18 and above. Google has said it plans to expand access later.

What are the limitations?

Google is upfront that Project Genie is still experimental. Current limitations include:

Worlds may not perfectly match prompts
Some character controls can feel limited
Sessions currently last up to 60 seconds
Advanced features like major world-changing events aren’t live yet

Feedback from early users will shape future updates.

The big picture

Project Genie arrives as competition heats up in generative AI. Companies like OpenAI, Meta, and World Labs are also investing heavily in spatial intelligence, AI systems that understand and generate 3D environments.

As DeepMind CEO Demis Hassabis wrote on X:
“Create entire playable worlds to explore in real time just from a simple text prompt—kind of mindblowing really.”

That just might be the understatement of the year.

Tech

ChatGPT to estimate users’ age to protect teens

OpenAI is rolling out a new safety feature in ChatGPT that aims to spot underage users — not by asking them to tick a box, but by predicting their age based on how they use the service.

According to TechCrunch, the company has introduced an “age prediction” system designed to identify accounts that likely belong to users under 18, and automatically apply stricter safeguards to limit exposure to sensitive content.

The feature works by analysing “behavioural and account-level signals,” TechCrunch reported, including details such as a user’s stated age, how long the account has existed, and patterns like the time of day the account is active.

If the model flags an account as belonging to someone under 18, ChatGPT will automatically shift that user into a more restricted experience, applying extra protections around topics such as sex, violence, and other material considered sensitive for minors.

Why OpenAI is making this change now

The move comes as pressure builds globally on tech platforms to strengthen protections for children and teens — especially as AI tools become part of daily life in schools and homes.

In a separate report, Reuters said OpenAI is rolling out age prediction globally as it prepares to introduce an “adult mode” for verified users in early 2026. Reuters added that users incorrectly flagged as under 18 will be able to regain full access by verifying their identity through a selfie submission to Persona, an identity verification service.

OpenAI has been building toward this approach for months. In an earlier policy update, the company described its long-term plan to tailor ChatGPT experiences depending on whether someone is over or under 18, including defaulting to safer protections when age is unclear.

The bigger trend: platforms moving to algorithmic age checks

OpenAI isn’t alone in using signals and automated tools to estimate a user’s age. Across social media, platforms have been ramping up similar systems, especially in markets with stricter online safety rules.

For example, The Guardian reported that TikTok has been strengthening age-verification technology across the EU using profile details, posted content, and behavioural signals to identify younger users.

For ChatGPT, the change reflects a growing shift in consumer AI: safety controls are moving away from static settings and into real-time systems that try to adapt to who is using the tool — and what they might be vulnerable to seeing.