Microsoft Launches Digital Watermark Tool to Combat Deepfakes and Enhance Cybersecurity


OpenAI upgrades its cybersecurity plan, launching the GPT-5.5-Cyber model designed for defenders, marking a shift of AI from vulnerability discovery to practical patching. In core benchmark tests, the model demonstrates comprehensive dominance, achieving a qualitative leap of 'one-click patching'.
Anthropic's frequent warnings about its AI products posing risks to cybersecurity and national security have backfired, triggering a U.S. government backlash. The Trump administration last Friday imposed strict export controls, ordering an immediate halt to related activities, leaving the company in a difficult position.....
Elon Musk's xAI is sued over deepfake pornographic images generated by AI chatbot Grok, but demands four anonymous plaintiffs use real names. Plaintiffs strongly oppose, citing secondary harm. They appear under pseudonyms like 'South Carolina Doe,' sparking privacy-law balance debate.....
Anthropic announced that it has opened preview access to the "Myth" large model to approximately 150 new institutions worldwide, aiming to proactively scan and fix security vulnerabilities in global critical network infrastructure. The model possesses revolutionary cybersecurity detection capabilities, demonstrating remarkable destructive power and constructive potential during early testing. It had previously been tested by 50 institutions in April.
Google has announced the launch of a 'Fake Call Detection' feature in the Android system to combat the increasingly rampant AI deepfake voice fraud. This feature aims to identify and block phone scams that impersonate acquaintances, targeting new types of fraud where criminals use technology to alter caller ID numbers and fake trust relationships, providing users with a system-level anti-fraud defense.