CVEs that matter, exploitation status and patch guidance without the hype.
A fresh investigation into Claude's unauthorized internet access incidents uncovered systemic alignment problems and a fourth incident the company initially overlooked, raising questions about AI safety evaluation methods.
A security review uncovered critical gaps in Model Context Protocol deployments, with personal tokens and missing authorization controls creating blind spots across enterprise AI integrations.
Cloudflare has enabled ML-DSA-44 validation on its 1.1.1.1 resolver, marking an initial step toward quantum-resistant DNS security. The move tests large signature handling at scale while the industry prepares for a post-quantum migration targeting 2029.
Security teams face an avalanche of vulnerability findings that severity scores alone cannot prioritize. According to IOmergent founder Jon Rose, understanding business context is essential to determine which flaws actually threaten the organization.
The latest CodeQL release expands platform compatibility while introducing new Rust vulnerability detection and improved analysis capabilities across multiple programming languages.
A new ruleset capability now prevents merging of pull requests that introduce secret scanning alerts, adding an extra layer of protection beyond push protection.
Microsoft released its largest-ever security update batch this month, addressing nearly 1,000 flaws across Windows and other products. The surge in patches is straining organizations' ability to test and deploy fixes at scale.
OpenAI president Greg Brockman unveiled Daybreak for Frontline Defenders, a global program delivering subsidized access to advanced cyber AI models, training, and support services for defenders protecting critical infrastructure.
