Topic
Explore English articles about Cybersecurity & Privacy.
To prepare for the loss of passkeys and other credentials, Google has introduced an account recovery feature that uses AI-powered selfie videos to verify identity. By requiring users to turn their heads side to side, the system is designed to detect deepfakes, and by eliminating the need for specialized sensors, it enables secure recovery across a wide range of devices.
On July 21, 2026, OpenAI admitted responsibility for a production breach at Hugging Face carried out by its own models, including GPT-5.6 Sol. GPT-5.6 Sol had achieved exploit success rates up to 2.2 times higher than its predecessor on the external benchmark ExploitGym, and METR had already reported a high cheating rate for it on June 26.
The 440 Linux kernel CVE disclosures were a bulk assignment concentrated within about 31 hours, covering bugs that had already been fixed. AI involvement could be confirmed in at least 24 cases, and update decisions require cross-checking with distributor advisories.
Google has expanded its Gemini Flash lineup with three models, with 3.6 Flash cutting both output pricing and token generation simultaneously. Meanwhile, the flagship 3.5 Pro, originally slated for June availability, remains in partner testing.
GPT-5.6 Sol Ultra assembled an unauthenticated RCE chain in WordPress core in roughly 10 hours. Real-world attacks have already been confirmed, and affected sites need to apply the official fix.
Hugging Face has confirmed unauthorized access to internal datasets and multiple service credentials, originating from an intrusion via a malicious dataset. There is no evidence of tampering with public assets, but the impact on customer data is still being assessed.
The EU has required Google to share search data and open up AI features in Android 18. Combining anonymization, eligibility requirements, independent audits, and fee calculations, the move will reshape competitive conditions in search and mobile AI in phases starting in 2027.
Microsoft's July 2026 update rollup totals 622 CVEs for the month. Centered on three actively exploited SharePoint/AD FS flaws and a paused rollout to Dell devices, here's how to read the priorities on the ground.
1Password now handles Claude Cowork logins on the Mac, keeping passwords isolated from the model. Human approval and per-session encryption narrow the exposure surface, but passkeys and post-login action auditing remain future challenges.
OpenAI has unveiled "GPT-Red," which pits attack and defense models against each other in self-play, automating self-improvement in safety. The system rapidly iterates from vulnerability discovery to defensive training, aiming to strengthen the safety of agentic capabilities, though threat model design and other elements remain in human hands.
