The IT Nerd

Attackers recently claimed to have breached T-Mobile, America’s second-largest mobile carrier, posting a dataset on a popular data leak forum. The post alleges that 64 million customer records were stolen as recently as June 1, 2025.

The Cybernews research team analyzed the data sample attached to the post on the forum to verify the nature and sensitivity of the leaked information.

The analysis found that the dataset contains a trove of highly sensitive user details:

• Full names
• Dates of birth
• Tax IDs
• Full addresses
• Phone numbers
• Email addresses
• Device IDs
• Cookie IDs
• IP addresses

If the data in the leak were new, researchers believe that individuals impacted face serious privacy concerns.

Leaked sample shows old and new data

T-Mobile has experienced multiple data breaches in recent years. The company previously agreed to pay over $15 million in fines for four separate incidents, two of which compromised the data of tens of millions of users.

Moreover, the Cybernews personal data leak checker indicates that at least some of the emails in the data sample were included in previous T-Mobile breaches.

However, certain fields — such as phone numbers — appear to be new, increasing concerns that this may represent a fresh breach. The full scale and authenticity of the 64 million records have yet to be confirmed with 100% accuracy.

To read the full research report, please click here.

Cybernews has uncovered a major data leak tied to Headero — a hookup app popular in queer and alternative dating circles — leaking over 4 million private records, including real-time GPS locations, sexual preferences, and explicit chats.

Over four million private records were found unsecured, including explicit chat logs, group messages, and detailed profile information, such as STD status and sexual preferences.

The most alarming part is that users’ precise GPS locations were also exposed, posing serious risks to personal safety, particularly for individuals in vulnerable communities.

How much data did the Headero app leak?

• 352,081 user records
• 3,032,001 chat records
• 1,096,904 chat room records

What data did the Headero app expose?

• Names 
• Emails 
• Social login IDs 
• JWT tokens
• Profile pictures 
• Device tokens 
• Exact GPS locations 
• Sexual preferences
• STD status

To read the full research report, please click here.

The Cybernews research team has uncovered a data exposure involving GonnaOrder, a Europe-based food delivery and digital ordering platform. The misconfigured system exposed thousands of customers’ personal details, including food orders, home addresses, and contact information.

Most of the affected users are located in the UK, Belgium, Greece, Germany, and the Netherlands. The team believes that the leaky instance has been open since August 2022, due to how it was indexed on an IoT search engine.

“Throughout the whole time the exposed instance was open, malicious actors could have obtained millions of customers’ data, including names, phone numbers, home addresses, as well as order details, which can often contain private info such as access codes to enter the building,” the research team said.

What data was exposed?

• Customer orders
• Restaurant and hotels where orders were made
• Customer phone numbers
• Email addresses
• Home addresses
• Delivery notes
• Payment methods used

What are the potential risks?

Exposed data can be used for identity theft, or sold on the dark web. Access codes in delivery notes may even enable physical crimes like burglary.

To read the full research report, please click here.

The Cybernews research team, in collaboration with cybersecurity researcher Bob Dyachenko, has uncovered a major data leak in China. A huge, 631 gigabytes-strong database was left without a password, publicizing mind-boggling 4 billion records.

The database consisted of numerous collections, containing from half a million to over 800 million records from various sources. The Cybernews research team believes the dataset was meticulously gathered and maintained for building comprehensive behavioral, economic, and social profiles of nearly any Chinese citizen.

Key research takeaways:

The database was quickly taken down, preventing researchers from identifying its owners. Before removal, the team accessed 16 separate data collections:

• 805 million records in a dataset labeled “wechatid_db” — likely user identifiers from WeChat, China’s largest messaging and payments app.
• 780 million records in “address_db” — containing residential information with geographic data, potentially revealing where users live.
• 630 million records in “bank” — included financial data, such as names, phone numbers, payment card numbers, and birth dates.
• 610 million records in a dataset translated as “three-factor checks” — likely tied to real-name identification, combining user IDs, phone numbers, and usernames.
• 577 million records in “wechatinfo” — believed to contain WeChat metadata, communication logs, or conversations.
• 300 million records in “zfbkt_db” — related to Alipay, containing card and token information; plus 20 million more records in a separate dataset with additional Alipay-linked financial data.
• 353 million records were spread across nine more collections, covering topics such as gambling history, vehicle registration, employment, pension funds, and insurance data.
• One dataset, tw_db, appears to focus on Taiwan-specific information, raising concerns of potential geopolitical surveillance.

Leak significance

The implications of this breach are far-reaching. Skilled attackers could identify where individuals live, their financial habits, debts, and even savings. With such comprehensive personal data, threat actors could launch highly targeted phishing attacks, commit identity or insurance fraud, or even engage in coercion or blackmail.

This lack of attribution means that affected individuals have no way of knowing their data was compromised, nor any channel to seek remediation.

To read the full research report, please click here.

Cybernews researchers have uncovered a major data breach involving the US-based dental marketing firm Gargle with a company leaking around 2.7 million patient profiles and 8.8 million appointment records online.

Key findings:

• The leak was caused by an unsecured MongoDB database. The data owner hasn’t been officially confirmed, but clues buried in the database point toward Gargle. 
• The leak likely stemmed from a common and often overlooked vulnerability where databases are left exposed without proper authentication due to human error. 
• Clues point to dental marketing company Gargle as the source, raising serious questions about third-party data handling and HIPAA compliance.

What data was leaked?

• Names
• Dates of birth 
• Emails
• Addresses
• Phone numbers
• Gender
• Chart IDs
• Language preferences
• Billing details
• Appointment records with patient metadata, timestamps, and institutional references

What are the potential dangers of this leak?

The leaked database includes highly sensitive patient details. When combined, this information creates a full blueprint of a person’s identity — opening the door to insurance fraud, medical identity theft, and highly targeted phishing and social engineering attacks.

To read the full research report, please click here.

Cybernews has discovered a major data leak with the world’s largest healthcare cooperative, Unimed, exposing 14 million patient-doctor messages. The data included uploaded pictures, documents, and other personal information.

What details are involved in the Unimed data leak?

• Uploaded pictures
• Uploaded documents
• Sent messages
• Names
• Phone numbers
• Email addresses
• Unimed card numbers

What are the potential dangers of this leak?

Healthcare data is highly valuable to cybercriminals, enabling identity theft, insurance fraud, phishing, and even blackmail. In this case, the breach was especially severe, as it could have allowed attackers to send, delete, or alter messages to users — opening the door to serious manipulation.

To read the full research report, please click here.

The Cybernews research team has uncovered a major data breach affecting 1.6 million customers of Etsy, Poshmark, and other TikTok shops, primarily in the U.S., with some affected users in Canada and Australia.

Two exposed instances revealed shipping confirmation emails in HTML format, exposing personal information such as full names and addresses.

What data was leaked? 

• Full names
• Home addresses
• Email addresses
• Shipping order details

Why is an Etsy shipping email leak dangerous?

• Attackers could impersonate Etsy or shipping providers to launch convincing phishing campaigns.
• Leaked order details make fraudulent emails appear legitimate, increasing the success rate of scams.
• Access to email and shipping info enables social engineering tactics to extract additional personal or financial data from victims.

To read the full research report, please click here.

Cybernews researchers evaluated the cybersecurity postures of the top 10 large language model (LLM) providers—including OpenAI, Claude, Perplexity, and DeepSeek — and discovered that half had suffered data breaches, with one breach occurring just nine days before the audit.

The Cybernews Business Digital Index, which evaluates companies based on key cybersecurity criteria, also revealed that all providers had vulnerabilities in their SSL/TLS configurations, and several faced widespread issues in system hosting, credential hygiene, and password reuse.

Additionally, nearly half of sensitive AI prompts are submitted via personal accounts, bypassing official company channels and potentially exposing corporate data without oversight — a growing risk as LLM tools become standard in the workplace.

Key research takeaways:

• 50% of the top LLM providers have experienced data breaches, including OpenAI (1,140 incidents) and Perplexity AI (190 credentials leaked just 13 days before the audit).
• All providers had SSL/TLS configuration issues, exposing them to potential man-in-the-middle attacks and data interception.
• Credential reuse was widespread — 35% of Perplexity AI employees and 33% of EleutherAI reused breached passwords.
• System hosting vulnerabilities were found in 8 out of 10 providers. Only AI21 Labs and Anthropic avoided major issues in this area.
• The average cybersecurity score across all providers was 88/100 — but scores ranged widely, with Inflection AI receiving an F.
• U.S. and Israeli providers generally scored higher than Chinese providers—none of the Chinese companies rated above a C.
• The growing use of personal accounts to interact with LLMs increases the risk of unmanaged data exposure.

To read the full research, please click here.

Research Methodology

For this study, Cybernews researchers analyzed 10 popular LLM providers. The report evaluates cybersecurity risk across seven key dimensions: software patching, web application security, email protection, system reputation, hosting infrastructure, SSL/TLS configuration, and data breach history.

The report’s Methodology can be found here. It provides detailed information on how researchers conducted this analysis.

The Cybernews research team has discovered that the iOS sleep management app Sleep Journey: Insomnia Helper exposed over 25,000 users via a misconfigured Firebase database.

The leaked data revealed names, email addresses, dates of birth, gender, sleeping data, habits such as alcohol and nicotine consumption, before-sleep activities, and medication use.

Here’s why this story matters:

• Leaking sensitive information like sleep patterns, substance use, and medical habits alongside names and emails gives cybercriminals everything they need to launch highly personalized and targeted attacks.
• It’s not just user data — it’s app infrastructure. Secrets like API keys, client IDs, and storage bucket credentials were hardcoded into the iOS app, potentially giving attackers high-level access to backend systems and user devices.
• Cybercriminals could use automated scrapers to harvest sensitive data in real-time — putting user privacy and service integrity at serious risk.
• It’s part of a larger, systemic issue. This breach was uncovered during Cybernews’ investigation into 156,000 iOS apps — revealing that 71% leak at least one secret, showing just how widespread insecure coding practices are.

To read the full research, please click here.

A junior developer at a fast-growing fintech startup, racing to meet a launch deadline, copied an API key into a public GitHub repo. Within hours, the key was scraped, bundled with others, and traded on Discord to a shadowy network of digital joyriders. 

By the time the company’s CTO noticed the spike in usage, the damage was done: thousands of dollars in LLM compute costs, and a trove of confidential business data potentially exposed to the world.

I’m not hypothesizing. It’s a composite of what’s repeatedly happened in the first half of 2025.

In January, the AI world was rocked by breaches that feel less like the old “oops, someone left a database open” and more like a new genre of cyberattack. DeepSeek, a buzzy new LLM from China, had its keys stolen and saw 2 billion tokens vanish into the ether, used by attackers for who-knows-what. 

A few weeks later, OmniGPT, a widely used AI chatbot aggregator that connects users to multiple LLMs, suffered a major breach, exposing over 34 million user messages and thousands of API keys to the public. 

If you’re trusting these machines with your data, you’re now watching them betray that trust in real time.

The New Playbook: Steal the Mind, Not Just the Data

For years, we’ve worried about hackers stealing files or holding data for ransom. But LLM hijacking is something different – something weirder and more insidious. Attackers are after the very “brains” that power your apps, your research, your business. 

They are scraping GitHub, scanning cloud configs, even dumpster-diving in Slack channels for exposed API keys. Once they find one, they can spin up shadow networks, resell access, extract more information for lateral movement or simply run up service bills that would make any CFO faint. 

Take the DeepSeek case, where attackers used reverse proxies to cover their tracks, letting dozens of bad actors exploit the same stolen keys undetected. The result? You could wake up to a massive bill for unauthorized AI usage – and the nightmare scenario of your private data, whether personal or professional, being leaked across the internet.

But the plot thickens with system prompt leakage. System prompts – the secret scripts that tell a GPT how to behave – are supposed to be hidden from the end users. But with the right prompt, attackers can coax models into revealing these instructions, exposing the logic, rules, and sometimes even extremely sensitive information that keep your AI in check. Suddenly, the AI you thought you understood is playing by someone else’s rules.

Why This Should Scare Us All

We’re wiring LLMs into everything, everywhere, all at once. Customer service bots, healthcare, legal research, even the systems that write our code. With every new integration, the attack surface grows. But our security culture might be still stuck in the times of password123.

In the meantime, the underground market for LLM exploits is exploding. Stolen keys are traded on Discord like baseball cards. Prompt leakage tools are getting more sophisticated. Hackers are sprinting ahead. And the more autonomy we give these models, the more damage a breach can do. We’re in a battle for control, trust, and the very nature of automation.

Are We Moving Too Fast for Our Own Good?

Thinking of AI as “just another tool” is a mistake. You can’t just plug these systems in and hope to slap on security later, because LLMs aren’t predictable spreadsheets or file servers. They’re dynamic and increasingly autonomous – sometimes making decisions in ways even their creators can’t fully explain. 

Yet, in the hurry to ride the AI gold rush, most organizations are betting their futures on systems they barely understand, let alone know how to defend. Security has been left in the dust, and the cost of that gamble is only going up as LLMs get embedded deeper into everything from business operations to healthcare and finance.

If we don’t change course, we’re headed for a reckoning – lost dollars and, more importantly, trust. The next phase of AI adoption will depend on whether people believe these systems are safe, reliable, and worthy of the power we’re handing them. If we keep treating LLMs like black boxes, we’re inviting disaster.

What Needs to Change, Ideally, Yesterday

So, what do we do? Here’s my take:

• Treat API keys like plutonium. Rotate them, restrict their scope, and keep them out of your codebase, chats and logs. If you’re still pasting keys into Slack, you’re asking for trouble.
• Watch everything. Set up real-time monitoring for LLM usage. If your AI starts unexpectedly churning out tokens at 3 a.m., you want to know before your cloud bill explodes.
• Don’t trust the model’s built-in guardrails. Add your own layers – filter user inputs and system outputs, always assume someone will try to trick your AI if it’s exposed to user input.
• Red-team your own AI solutions. Try to break it before someone else does. 
• Implement segregation through access controls. Don’t let your chatbot have the keys to your entire kingdom.

And yes, a handful of vendors are starting to take these threats seriously. Platforms like Nexos.ai offer centralized monitoring and guardrails for LLM activity, while WhyLabs and Lasso Security are developing tools to detect prompt injection and model emerging threats. None of these solutions are perfect, but together they signal a much-needed shift toward building real security into the generative AI ecosystem.

Your AI’s Brain Is Up for Grabs, Unless You Fight Back

It’s time to recognize that LLM hijacking and system prompt leakage aren’t sci-fi. This stuff is happening right now, and the next breach could be yours. AI is the new brain of your business, and if you’re not protecting it, someone else will take it for a joyride.

I’ve seen enough to know that “hope” isn’t a security strategy. The future of AI seems bright, but only if we get serious about its dark side now – before the next breach turns your optimism into regret.

ABOUT THE AUTHOR

Vincentas Baubonis is an expert in Full-Stack Software Development and Web App Security, with a specialized focus on identifying and mitigating critical vulnerabilities in IoT, hardware hacking, and organizational penetration testing. As Head of Security Research at Cybernews, he leads a team that has uncovered significant privacy and security issues affecting high-profile organizations and platforms such as NASA, Google Play, and PayPal. Under his leadership, the Cybernews team conducts over 7,000 pieces of research annually, publishing more than 600 studies each year that provide consumers and businesses with actionable insights on data security risks.