Security
Security Flash: Quantum Risks, AI Leaks, Europol Overhaul
Today's top stories: China's quantum computer test, Anthropic model shutdown, Telegram ban upheld, Europol shadow IT, and more.
China’s Wukong Quantum Computer Runs PQC Shield Amid National Security Concerns
On June 17, the 72-qubit Origin Wukong superconducting quantum computer completed over one million tasks under an integrated post-quantum cryptography (PQC) framework. The system, operated by Origin Quantum in Hefei, has been stable for over two years and received over 49 million remote visits from users in 192 countries. Its software cryptographic module, Origin Rock, protects data from ‘harvest now, decrypt later’ attacks, but experts warn that the shield cannot protect against compelled disclosure under China’s National Intelligence Law. Origin Quantum is on the US Entity List, and no independent security audit of Origin Rock has been published. The Wukong chip poses no threat to breaking RSA-2048, requiring roughly 1,400 logical qubits versus its 72 physical qubits. NIST finalized three PQC standards in August 2024, while China is developing its own standards with less international scrutiny, expected within approximately three years.
Post-Quantum Cryptography Meets China’s Quantum Computer: What the Shield Leaves Exposed →
AI Chatbots Pose Data Leak Risks; Third-Party Exposure of 300 Million Messages Reported
Cybersecurity professionals in Chicago warned that pasting sensitive personal information into AI chatbots can lead to data leaks. Check Point Research documented a DNS-tunneling side channel in ChatGPT’s code-execution runtime, which OpenAI fixed on Feb. 20, 2026. Additionally, a widely used third-party AI chat app exposed roughly 300 million messages from over 25 million users due to a misconfigured Firebase instance. Researchers also flagged malicious Chrome extensions that exfiltrate conversation content via ‘prompt poaching.’ Experts advise never sharing Social Security numbers, bank details, passwords, API keys, or medical records into public AI tools. OpenAI offers opt-out options for model training, but users should review documentation before sending sensitive information.
Chicago Experts Warn AI Chatbots Could Leak Data →
Anthropic Shuts Down Two AI Models After US Flagged SK Telecom’s Alleged China Links
Anthropic took its Claude Mythos and Fable 5 models offline after the White House expressed alarm over SK Telecom’s alleged ties to China. SK Telecom had access to Mythos through Anthropic’s partner program Project Glasswing. The White House ordered Anthropic to cut off SK Telecom’s access, and Anthropic complied immediately. Separately, Amazon and other companies flagged security flaws in Fable 5 that could bypass safety restrictions. The two incidents combined led the White House to lose confidence in Anthropic, forcing both models offline entirely.
Alleged China ties at SK Telecom alarmed US officials and triggered Anthropic crisis →
Google Deepmind Rolls Out AI Control Roadmap Treating Agents as Insider Threats
Google Deepmind developed an AI Control Roadmap that treats internal AI agents as potential insider threats, assuming they may not share operators’ goals. The framework builds on MITRE ATT&CK to track risks and spot suspicious behavior. Deepmind built an internal prototype monitoring system for coding agents, running it across one million tasks. The system already monitors the Gemini Spark agent live, catching issues like accidental data deletion. Deepmind also published a paper on ‘Three Layers of Agent Security’ and warned that there is a narrow window to lock in security protocols before multi-agent systems scale globally.
Google Deepmind treats its own AI agents like rogue employees with office keys →
Europol’s Shadow IT Database Exposed; Brussels Proposes Doubling Budget and Powers
A May 2025 investigation uncovered that Europol maintained a parallel ‘shadow IT’ database, at one point hosting over 95% of its collected data and being 420 times larger than official databases, operating outside legal safeguards. The European Commission is expected to propose an overhaul on June 24, 2025, that would double Europol’s budget to €444 million by 2034, expand its staff, remove data protection safeguards, and weaken EDPS oversight. Europol processes data on political activities and travel, contributing to surveillance and criminalization of social movements. Affected individuals are rarely informed about data sharing with foreign authorities.
Europol is going rogue – so Brussels is doubling its budget?? →
Singapore Startup Brings Diophantine Equation Encryption to Market with LionGuard App
Lim Meng Liang and Ken Lin co-founded Aires Applied Quantum Technology in 2023 to commercialize encryption based on unsolvable Diophantine equations. Their flagship product, LionGuard, is a mobile app that encrypts files on devices, cloud, or networks, costing $388 per user per month. The app is in beta with over 100 subscribers across oil and gas, banking, and other sectors. Aires holds four international patents and has raised over US$2 million. The brothers aim to lower subscription fees to attract SMEs and are seeking to list in the US, Singapore, or Japan for global expansion.
Singaporean brothers use unsolvable maths equations to build modern, unbreakable encryption →
Delhi High Court Upholds Full-Platform Telegram Ban Over NEET Exam Fraud
The Delhi High Court ruled on June 19 that the government can block an entire messaging platform under Section 69A, dismissing Telegram’s challenge to a nationwide ban that cut off over 150 million Indian users. The ban was imposed after criminal networks used Telegram channels to defraud medical school aspirants around the NEET-UG 2026 exam. The court accepted the government’s argument that Telegram’s architecture made targeted enforcement structurally impossible. The ruling also upheld a separate order requiring Telegram to disable its message-editing feature for Indian users through June 30. The Internet Freedom Foundation called the ban disproportionate and warned it sets a precedent for other encrypted apps like Signal and WhatsApp.
India’s Telegram Ban Upheld by Court, Putting 150 Million Users and Every Encrypted App at Risk →
ShinyHunters Publishes 297GB of Council of Europe Data After Ransom Deadline Missed
The criminal group ShinyHunters released 297 gigabytes of Council of Europe employee data after the organization failed to meet a June 16 ransom deadline. The breach, enabled by a critical Oracle PeopleSoft zero-day (CVE-2026-35273), exposed over 10,000 current and former employees’ bank details, medical records, and salary histories spanning 15 years. ShinyHunters announced the data would be permanently available via mirrors and torrents, eliminating takedown options. The Council of Europe confirmed it is investigating but has not yet notified affected individuals or offered credit monitoring.
Council of Europe Data Breach: ShinyHunters Makes 10,000 Employees’ Records Permanent →
Alabama Defense Contractor LOGZONE Pays $507k for Missing Navy Cybersecurity Safeguards
LOGZONE Inc. agreed to a $507,144 settlement under the False Claims Act for failing to maintain required cybersecurity protocols on two U.S. Navy contracts from May 2021 to March 2025. A DCMA assessment found the company scored -170 on a scale from -203 to 110 for compliance with NIST SP 800-171. The Justice Department stated the failures could have allowed exploitation of sensitive defense information. The settlement is part of a broader federal crackdown on government contract fraud, involving multiple agencies including NCIS and DCMA.
Failed Cyber Test: Alabama Defense Contractor Settles For $500K Over Missing Navy Safeguards →
Fraunhofer IPMS Unveils 4.1 Gb/s Quantum Random Number Generator for Data Centers
Fraunhofer IPMS introduced Q-Dice, a quantum random number generator that uses quantum vacuum fluctuations to produce randomness at 4.1 Gb/s. The system is a 19-inch rack-mounted appliance with 10 Gbit/s Ethernet, designed to provide entropy that cannot be predicted or reverse-engineered. It has been evaluated with BSI AIS 20/31 and NIST SP 800-22 tests and carries EAL 3 and PTG 3 classifications. Fraunhofer is seeking partners to pilot the technology and also offers an Entropy-as-a-Service platform.
Over 500 Billion Monthly DNS Queries to Residential Proxy Networks Threaten Businesses
Infoblox Threat Intel found that over 65% of its Threat Defense Cloud customers made DNS queries to domains linked to residential proxy networks in 2026, totaling over 500 billion such queries monthly. These proxies route traffic through ordinary home routers and IoT devices, making malicious traffic appear as normal consumer activity. Many free VPNs and streaming apps rent out user network identity without explicit permission, often through buried EULA consent. Infoblox recommends software audits, router-level blocking, and protective DNS to mitigate risk.