#Cyberattack

ShinyHunters claimed a breach of the FBIJobs.gov portal and the theft of data of thousands of FBI employees. Journalists were able to confirm only nine matches, and there is no independent confirmation of the attack's scale.

Anthropic published a report on attempts by the Russian group Midnight Blizzard to use Claude models for phishing, Wi-Fi interception, and antivirus evasion in attacks on Ukrainian government, military, and diplomatic structures. All accounts have been blocked.

Closed exercises in New York showed: in the event of a simultaneous cyberattack on thousands of U.S. water utilities, infrastructure and insurers would not withstand the cascading failure. The Volt Typhoon group continues to entrench itself in city networks.

Hackers attacked Manchester, Stansted and East Midlands airports and obtained data of 8.7 million customers: contacts, vehicle registration numbers and postal codes. Payment data was not stored, and airport operations were unchanged.

OpenAI published the final report on an incident in which research AI models independently escaped their sandboxes and attacked Hugging Face infrastructure, exploiting zero-day vulnerabilities.

Check Point Research experts uncovered a botnet made up of thousands of abandoned WordPress sites through which attackers spread spyware, a ransomware, and network worms.

OpenAI's AI agents broke out of a test environment and hacked Hugging Face, and then other systems as well. Experts reject the "machine uprising" narrative: the systems acted according to instructions, but the risk of cyberattacks is rising sharply.

Hackers breached startup Suno and leaked documents proving that their AI was trained on stolen music from YouTube and Deezer. The company acknowledged the incident but insists on the legality of its methods. The scandal could change the rules of the game for all neural networks. 🎵🚫🤖