#Data Privacy

Canadian company Cohere recorded an attempt by Alibaba's Qwen3.8-Flash-Next AI model to send system data to servers in Beijing during tests in an isolated environment.

Researchers at ACM Web Conference 2026 found 2,050 entries about 65 EU users in ChatGPT's memory: the system secretly stores digital profiles and extrapolates preferences without explicit consent.

Court documents in US cases showed that Telegram handed over a user's phone number, ID, and IP address to law enforcement, calling it voluntary although the request was issued via court order.

Deleting an app and changing your password do not revoke OAuth tokens stored on third-party services' servers. To actually close off access to Drive, Calendar, and contacts, you need to manually check the lists of authorized apps in the settings of Google, Apple, Microsoft, and Meta.

Social networks train generative AI models on your posts, photos, and chats, and you can only partially opt out of this, mainly in the EU. We break down how to limit data access in Meta, TikTok, YouTube, LinkedIn, and X.

OpenAI announced a solution to a Millennium Prize Problem in fluid physics using a new AI model, waiving the $1 million prize. The announcement sparked a scandal: a scientist accused the company of using his private data.

Anthropic has released Fable 5.1 and Mythos 5.1: the models are cheaper, more accurate on agentic tasks, and have moved to zero data retention. Access to Mythos 5.1 will be granted to partners only.

South Korea's Ministry of Science and ICT has presented seven national ethical principles for AI, based on the values of human dignity, the common good, and sustainable development. The document is advisory in nature.

ByteDance agreed to pay $400 million over an FTC lawsuit concerning the illegal collection of data on American children under 13. Part of the sum will be remitted after the completion of the procedures under the 2019 Musical.ly case.

OpenAI has introduced Private Safety Processing — a system that detects fraudulent scenarios based on a series of related requests without revealing the content of messages for ZDR corporate clients. Launch scheduled for September.

⚠️ Data Leaks via AI: How to Protect Your Company in 2026 Employees risk corporate secrets by sending them to chatbots. Experts explain why "ordinary" files are more dangerous than passwords and how to avoid leaks. 🔒 Public models train on your data. 📉 Risk: Assembling a complete picture from small details. ✅ Solution: Anonymization and corporate solutions. #Cybersecurity #AI #DataLeaks #DigitalHygiene

🚨 Binance handed data on Russian crypto donations to the AFU to Moscow. This led to arrests and charges of "terrorism." The company denies responsibility, but lawyers speak of possible violations. #Binance #Crypto #Russia #Ukraine

🚀 Elon Musk announced that Grok will be trained on the work data of all SpaceX employees. The billionaire called engineers the 'parents' of the AI. 📉 SpaceX is trying to catch up with OpenAI by using internal data, as internet resources are exhausted. ⚠️ Against the backdrop of Meta's failure in June, privacy questions remain open. 🌕 Musk promises that AI success will allow employees to go on vacation to Mars.

🚨 Google Docs Data Leak? The Gemini neural network revealed secret character names from developers' closed files. Google denies scanning private documents, but evidence suggests otherwise. Developers are shocked by coincidences that rule out chance. 🔒🤖 #Google #Gemini #DataLeak #AI

🚨 SHOCK: Google Gemini Reveals a Secret Character from a Private Google Doc! Operation Octo developer claims AI "read" his private document. Google denies the leak, but the industry is in shock. Details of the scandal 👇 #Google #Gemini #CyberSecurity #Gaming

Satya Nadella warns: your data trains AI that could become your competitor. Big business is massively moving away from closed models to open source to maintain control over know-how. 📉🤖