OpenAI Launches Patch the Planet to Help Open Source Maintainers Fix Vulnerabilities
OpenAI Launches Patch the Planet to Help Open Source Maintainers Fix Vulnerabilities
OpenAI introduces Patch the Planet, a new initiative under the Daybreak program. It aims to assist open-source maintainers in finding, validating, and fixing vulnerabilities using AI and expert review. The project addresses the resource scarcity in open-source maintenance by offering automated vulnerability detection and remediation suggestions alongside a human review process.
OpenAI Board Member & Gray Swan CEO: Why AI Security Is Not Just Cybersecurity
In this episode of Latent Space, OpenAI board member Zico Kolter and Gray Swan CEO Matt Fredrikson join swyx to discuss AI security and Red Teaming. They argue that AI security is fundamentally different from traditional cybersecurity and highlight the blind spots in current testing methods. The conversation covers the need for more robust safety evaluation frameworks and why existing defense mechanisms fail against specific attack patterns.
Prompt Injection Explained as Role Confusion: System Instructions Can Be Spoofed
Simon Willison provides a blog-style breakdown of a new paper characterizing prompt injection as role confusion. The research highlights that LLMs struggle to distinguish between user input and developer instructions. When system prompts and user data are mixed, models experience role confusion, allowing attackers to override safety instructions by spoofing data.
Porting Moebius 0.2B Inpainting Model to Browser with Claude Code
Simon Willison demonstrates porting the Moebius 0.2B image inpainting model to run in the browser using Claude Code. Moebius is a lightweight framework with performance comparable to 10B-level models but originally required a Python environment. With Claude Code, Willison generated WebAssembly-compatible code, enabling in-browser image inpainting.
Selector Forge: Open Source AI Extension Generates Resilient CSS/XPath Selectors
The Intuned team has released Selector Forge, an open-source AI browser extension. It leverages AI to automatically generate resilient CSS or XPath selectors for web scraping or automation, solving the issue of selector fragility.
Replit's First Vibecon Highlights Rise of 'Vibe Coding' Developer Trend
Replit hosted its inaugural Vibecon event in New York’s Lower East Side, transforming an industrial space into a vibrant hub for developers. The event highlighted the growing popularity of Vibe Coding, showcasing Replit’s AI-assisted coding environment. It demonstrates the shift from traditional IDEs to collaborative AI-driven development experiences.
Meta Pauses Employee Keystroke Tracking AI Program After Internal Leak
According to Business Insider, Meta has paused an internal AI training program that tracked employee keystrokes after the initiative was leaked and sparked controversy. The program aimed to collect data from work devices to train internal AI models. Following privacy concerns and employee backlash, Meta halted the data collection.
GitHub Updates Progress on Inclusive Open Source and Accessibility Pledge
GitHub Blog reports on progress made toward accessibility goals and building a more inclusive open-source ecosystem. The post details new tools and resources designed to help developers with disabilities participate in open source. GitHub calls on the community to help implement these Pledge commitments through technical improvements and documentation updates.
Claude Code v2.1.186: CLI MCP Login and SSH Stdin Redirect Support
Claude Code released version v2.1.186 with new CLI features. It added claude mcp login and logout commands to authenticate MCP servers without the interactive menu. The update also supports --no-browser stdin redirection for SSH usage and introduced status filtering (press f) in the /workflows agent detail view.
OpenClaw 2026.6.10-beta.2: Auto Fast Mode and Improved Model Routing
OpenClaw released version 2026.6.10-beta.2. Key highlights include an automatic fast mode for short conversational turns, which reverts to normal mode for longer runs with bounded fallback. The update also features more reliable model routing via Zai model synthesis, improving performance balance across different conversation lengths.
OpenAI Codex Releases Preview Version 0.143.0-alpha.2
OpenAI Codex released version 0.143.0-alpha.2, following the 0.143.0-alpha.1 and 0.142.0 series. This is a routine pre-release iteration focusing on bug fixes and underlying optimizations.
OpenAI Launches Daybreak Security Tools with Codex and GPT-5.5-Cyber
OpenAI has launched the Daybreak suite of security tools, featuring Codex Security and GPT-5.5-Cyber. These tools are designed to help organizations find, validate, and patch vulnerabilities at scale.