DeepSeek User Conversations Appear in Google Search Results via Publicly Shared Links
Conversations between home users and the Chinese AI service DeepSeek, including many in Russian, have been discovered in Google search results. Accessing another person’s dialogue with the model requires only a specially crafted search query.
The pages in question are Shared Conversations — dialogues that users voluntarily published via public links. Every question, answer, and the titles of any uploaded documents are visible on these pages, as verified by a TASS correspondent. Journalists were unable to download the actual files, even by inspecting the page source code.
Representatives of MWS AI stated that no new data leak has taken place. They noted that search engines had already indexed thousands of publicly shared chatbot dialogues more than a year ago. No reports of a compromise of DeepSeek or unauthorized access to private conversation histories have emerged.
Most likely, the indexed pages are either old conversations or ones that users deliberately made public. It remains impossible to obtain access to a user’s account or private chats through these links. In effect, DeepSeek did not open doors to personal accounts; users themselves left certain conversations publicly accessible, and Google simply catalogued them.
In parallel, Yandex confirmed that links to dialogues with its Alice AI assistant are blocked from search-engine indexing. Such links are created only at the user’s initiative, remain valid for 14 days, and can be revoked at any time. Attached files are never transmitted to the recipient, and continuing another user’s conversation stays private.
Related articles
New Obfuscation Method Dissolves Personal Data Records in Layer of Plausible Variants
A Russian information security researcher has proposed a data protection technique that renders stolen personal records unusable even after full compromise. The approach mixes real data such as phone numbers, emails, passports, addresses, INN and SNILS with vast numbers of semantically valid alternatives. Attackers receive nearly complete information including a 361-character message containing PIN codes and word order, yet lack the secret vector space and reconstruction algorithm required to identify the correct record. Without these components, brute-force attempts produce millions of plausible results with no architectural method to verify accuracy. The method is presented as an alternative to traditional encryption when data must remain accessible yet protected against extraction. A public sandbox is available for testing the approach.
Hydrat Project Builds Automated WireGuard Gateway for Resilient VLESS and Tor Routing
A developer has released Hydrat, a self-hosted gateway that connects devices via WireGuard while automatically managing VLESS and Tor backends to survive server blocks and quality degradation. The system maintains a pool of tested proxies, performs continuous health checks, and switches routes without requiring client-side profile changes. Two Go processes handle control logic and network enforcement separately, using SQLite for state and nftables plus Xray for traffic routing. TCP and UDP can be assigned independent exits, with geoip.dat support and custom rules to keep marketplace apps functional. The project emphasizes stability over direct connections and is designed for deployment on servers in Russian jurisdiction.
OpenAI Contractors Manually Review Real User Chats in Project Lily
OpenAI has engaged hundreds of external contractors to analyze actual user conversations with ChatGPT as part of its model improvement efforts. The reviewers, working under project Lily, examine real queries that may contain personal, medical, or other sensitive information despite the use of a Privacy Filter. Contractors summarize prompts, compare four model responses, and assign ratings from one to seven while flagging behaviors such as excessive sycophancy or inappropriate emojis. User identities are hidden and some data is filtered, yet OpenAI acknowledged that not all personal information is reliably removed. The same human review process is also employed by Anthropic for its Claude model. Users can opt out of future training use through account settings, although prior data remains unaffected.
UDP Proxies and QUIC Protocol: How Real IP Addresses Leak Through Anti-Detect Browsers
Anti-detect browser users relying on UDP-capable proxies face a hidden risk of real IP leakage when the browser fails to properly route UDP traffic. The QUIC protocol, which powers HTTP/3, runs over UDP and enables features like 0-RTT handshakes, independent streams, and connection migration that can bypass proxy routes. WebRTC connections using ICE, STUN, and TURN further increase exposure because they often attempt direct UDP paths outside the configured SOCKS5 proxy. Without deep network stack control such as TUN interfaces or socket interception, browsers may send WebRTC and QUIC packets through the host's real network interface. Aurorium Browser claims to solve this by natively supporting UDP proxying so that both QUIC and WebRTC traffic stays inside the tunnel. The article stresses that simply disabling WebRTC or forcing HTTP/2 fallback is insufficient and can itself create detectable anomalies for anti-fraud systems.