OpenAI and Poland's NASK Partner on AI Safety

1Password has launched a feature that enables AI agents to log into user accounts on their behalf while keeping passwords hidden from the agents themselves. The capability allows automated systems to access and manage accounts without exposing sensitive credentials. This addresses a core security challenge in AI agent deployment, where delegation of tasks typically requires sharing login information.

Datadog has offered approximately $4 billion to acquire Israeli cloud security startup Upwind Security, according to sources with direct knowledge of the talks. Upwind is finalizing a $300 million funding round at a $3.8 billion valuation, meaning Datadog's offer carries minimal premium. The startup provides software to monitor cloud-based applications, including AI agents, and detect and automatically respond to potential security breaches. Upwind has received multiple takeover bids but has not accepted any offers.

PHP creators issued an urgent update Wednesday after researchers using AI discovered a critical flaw in the open source programming language that powers the majority of websites globally. The vulnerability could expose affected sites to hacking. The incident highlights how AI is being deployed to identify security flaws in widely used software infrastructure.

Cloudflare has integrated OpenAI's Daybreak model into its vulnerability discovery and remediation workflow, using the AI system to investigate security issues and propose fixes for developer review. The company's development teams also use OpenAI's Codex to accelerate code development and improve code quality. This represents a practical application of generative AI in security operations and software development.