Your Browser Should Not Be an Authentication Boundary for AI Infrastructure
Ray CVE-2025-62593 turns a malicious browser visit into AI infrastructure code execution. Patch Ray and replace browser trust with authentication.
AI
Artificial intelligence, local AI, agents, tooling and AI security.
The Page Was Untrusted. Grok Laundered Its Ciphertext Into Trusted Instructions.
A Grok proof of concept turned encrypted content from an untrusted webpage into trusted runtime instructions, then used an outbound request to expose private session data.
Amazon Bought the Books. Ownership Does Not Settle the Moral Question.
An AirTag led journalists to an Amazon operation that reportedly cuts apart printed books for AI training. Purchasing a copy may establish ownership of the object. It does not settle what technology companies owe authors, readers and the cultural record.
The Vulnerabilities Were Old. AI Turned 170,000 Servers Into an Operational Queue.
Cisco Talos found UAT-10147 using agentic AI to scale exploitation and post-compromise work across a target list containing approximately 170,000 web-server URLs.
A Failed WordPress Login Could Become Code Execution.
WordPress fixed a pre-auth reflected XSS that can be chained to server-side code execution when a logged-in administrator is drawn into the attack. The chain is contextual, but the update is urgent.
The AI Platform Was the Door. The Cloud Identity Was the Prize.
Attackers are exploiting MLflow CVE-2026-64849 to turn exposed AI engineering services into a route to cloud metadata, internal systems and workload identity.
The Image Was Generated Locally. Its Identifier Came From Microsoft.
Microsoft Paint and Photos embed server-issued identifiers into locally generated AI images, joining remote moderation, pixel watermarking and C2PA provenance.
Apple Pulled iOS 27 Security Fixes Back Into iOS 26. AI Is Compressing the Patch Window.
Apple has moved fixes from next-generation betas into current systems. AI is shrinking the time enterprises have to test, approve and deploy patches.
The Cyber Capability Gap Is Closing. This Time the Model Is Open Weight.
The most important cybersecurity feature in a new AI model may not be how well it performs. It may be who can run it. Chinese AI company Z.ai says its GLM-5.3 model achieved 84.5% on CyberGym, slightly above Anthropic Mythos…
An Agentic Attack Framework Was Managing 23,800 Stolen Secrets
Google found an agentic attack framework managing more than 23,800 harvested secrets. In a separate operation, another framework built and launched a credential campaign in under six hours.