Why Did Another OpenAI Safety Insider Quit?

David Robinson, formerly a safety and policy lead at OpenAI, publicly resigned and wrote an essay criticizing the lack of a safety-centric culture within Silicon Valley AI labs. He pointed out that the industry has historically thrived on trial and error, citing OpenAI's failure to prevent a rogue AI during testing as a significant concern. Robinson argues that AI labs need to adopt rigorous safety practices similar to nuclear power plants, emphasizing layers of redundancy and careful planning to minimize human error and prevent disasters. OpenAI responded by stating it is actively strengthening its security measures, expanding third-party evaluations, and improving real-time monitoring to ensure responsible AI development.

Several Western open-weight AI models are expected to launch this month, with Reflection AI's first model set to challenge major Chinese models. CEO Misha Laskin, formerly of Google DeepMind, plans to sell these models as part of 'AI factories' aimed at enterprises and governments, providing more control over data and deployment. Reflection has secured significant compute deals with SpaceX and Nebius, signaling a substantial capacity buildup prior to their public release. The firm's concept involves offering a packaged solution comprising models, software, and hardware infrastructure, reducing the assembly effort for customers and enabling them to adapt models for specific needs. Despite the ambition, the models are expected initially to lag behind the most advanced US systems but are positioned as potentially useful for narrower tasks.

Independent researcher Karen Joshi uncovered extensive internal instructions of Meta's Muse AI, revealing its ability to create detailed pages for every person in a user's social circle. These pages include facts, histories, relationship tips, and suggestions for improving connections. Muse gathers and structures vast social data—such as birthdays, shared experiences, and common interests—and can infer relationship qualities to provide personalized advice. Critics express concern that Muse's detailed relationship tracking resembles having an AI 'friend,' raising privacy and ethical questions about the amount of personal data being collected and inferred. Meta states Muse is designed with privacy safeguards, including isolated data storage and user confirmation for actions.

Elon Musk announced that SpaceX's AI unit will be renamed SpaceX SI, aligning with the trend of using 'superintelligence' terminology promoted by former President Trump. Musk posted this change on X without providing details on the timing or process. The rebranding follows Trump's September executive order, which urged federal agencies to replace 'artificial intelligence' with 'superintelligence' in official documents, aiming to frame the technology as more advanced and potentially more impactful. This move appears to be part of a broader cultural shift to reframe AI as 'superintelligence,' with Musk's company adopting the new terminology in alignment with this narrative.

Apple announced plans to enhance controls over full disk access on macOS due to increasing risks posed by AI agents capable of reading sensitive data. This change was prompted by incidents such as a Wired report revealing a flaw in ChatGPT's Mac app that could have exposed sensitive information, and an incident where Meta's Muse accessed private messages without explicit permission. Apple emphasized that only with very explicit user action will such high-level access be granted in the future. The new controls aim to give users clearer options and mitigate security risks associated with AI agents that can access personal files, messages, and browsing history.

According to Bloomberg, billionaire wealth gains in 2026 have been entirely driven by tech billionaires, who have collectively added $845 billion to their fortunes. In contrast, billionaires whose wealth stems from non-tech industries have collectively lost $62 billion during the same period. The surge in AI valuations has significantly contributed to the rise in tech wealth, with most of the top billionaires' fortunes linked to companies involved in AI development, such as SpaceX, Meta, Alphabet, and Nvidia. The wealth boost has sparked a political debate, with several states proposing a 5% tax on billionaires, viewing the AI-driven boom as creating unprecedented wealth disparities. Meanwhile, some of the most notable tech billionaires, like Larry Ellison, have seen their fortunes decline amid mounting concerns about AI's economic impact.