Legacy Code

OpenAI has published guidance on how data science teams can leverage ChatGPT Work to automate the generation of common analytical outputs: root-cause analysis briefs, impact assessments, KPI memos, scoped analyses, and dashboard specifications. The platform takes raw work inputs and generates structured, business-ready documents that typically required significant manual effort.

Gidi Littwin, a veteran of Apple's Face ID team, has founded Hemispheric, a startup developing AI systems to analyze brain scans for early detection of conditions including depression, PTSD, and Parkinson's disease. The company aims to make diagnostic brain imaging as accessible and affordable as a blood test, leveraging AI to extract clinical insights from existing imaging data.

Researchers have identified a systemic vulnerability in Microsoft's Secure Boot implementation that has persisted since the feature's introduction. The flaw stems from legacy compatibility "shims"—code patches designed to support older operating systems—that Microsoft failed to revoke or properly validate over the years. Attackers can exploit these overlooked shims to completely bypass Secure Boot's protections, allowing unauthorized code execution at the system level.

Microsoft released an unusually large number of security patches in a single update cycle, directly coinciding with the disclosure of a new 0-day vulnerability called HiveLegacy. Security researchers characterize HiveLegacy as a "powerful primitive"—meaning it provides a foundation for more complex attacks that could lead to system compromise or lateral movement across networks.

OpenAI has published a policy framework advocating for what it calls "reverse federalism" in AI governance—a model where state-level AI regulations help inform and shape national standards, rather than the opposite. The proposal seeks to establish baseline safety and transparency requirements while allowing states to experiment with sector-specific rules tailored to their economies and values.

OpenAI has launched a $230 light-up keyboard designed to work with Codex, its AI-powered coding assistant. The move marks the company's entry into hardware, though notably not the consumer device currently in development with designer Jony Ive—that project remains embroiled in a lawsuit with Apple over alleged trade secret theft.

OpenAI has unveiled GPT-Red, an automated red teaming system that uses self-play—essentially having an AI attack and defend against itself—to identify and fix security vulnerabilities in language models. The system was instrumental in hardening GPT-5.6, OpenAI's latest flagship model, improving its defenses against prompt injection attacks and other adversarial techniques.

Data obtained through a security breach has revealed that Suno, a popular AI music generation platform, trained its models on millions of copyrighted songs scraped from YouTube Music, Deezer, and other platforms without explicit licensing. The exposure raises serious questions about fair use claims and could fuel ongoing legal challenges against AI music companies by the recording industry.

Thinking Machines Lab has released Inkling, a 975-billion-parameter open-source AI model trained to process both video and audio alongside text. The model represents the company's first public offering after 18 months of building AI infrastructure largely out of public view, positioning itself as an alternative to the dominant one-size-fits-all approaches from competitors like OpenAI and Anthropic.