Legacy Code

Anthropic published research demonstrating that it can observe and interpret Claude's internal thought processes by examining the model's hidden computational layers. The research describes a 'global workspace' mechanism within Claude where the AI appears to work through concepts before generating responses, similar to human conscious reasoning.

Apple has filed a lawsuit against OpenAI, claiming that current and former employees stole confidential hardware specifications, prototype designs, and supplier information to advance OpenAI's own hardware initiatives. The complaint alleges that senior leadership at OpenAI directed the misconduct and that affected employees brought confidential presentations and technical documentation from Apple to their new roles.

Anthropic's research reveals that Claude includes a 'global workspace' architecture—a computational layer where the model consolidates and reasons through concepts before generating final responses. This structure is observable by researchers examining Claude's internal states, allowing them to trace the model's reasoning process step-by-step.

Deutsche Telekom has deployed OpenAI's technology across multiple operational domains, including customer service, employee workflows, network operations, and voice services. The telecommunications giant is using AI to become an 'AI-native' operator, fundamentally restructuring how it handles routine customer interactions and internal processes.

Instagram head Adam Mosseri stated in a recent podcast interview that the platform will not filter out AI-generated content, instead emphasizing user preference controls. His position suggests Instagram intends to keep AI content visible by default while allowing individuals to adjust their own feed settings to reduce or eliminate AI posts.

A report found that Meta's AI-generated content detection tool cannot reliably identify images that Meta's own AI systems created. This fundamental limitation undermines Meta's ability to transparently label AI-generated content on its platforms, making it harder for users to distinguish synthetic images from authentic ones.

Meta has disabled a newly launched Instagram feature that allowed users to generate AI images based on public posts from other users' accounts. The tool could create deepfake-style images simply by tagging any public Instagram account, without explicit consent from account owners. Following significant user and creator backlash over privacy and consent concerns, Meta decided to pull the feature entirely.

OpenAI is recruiting a dedicated product manager to develop ChatGPT experiences specifically designed for families, caregivers, and older adults. The job posting suggests the company sees significant market opportunity in adapting its flagship product for demographic groups that may have distinct needs and usage patterns.

Johannes Heidecke, OpenAI's Head of Safety, is leaving the company as it consolidates its research and safety teams into a single organizational unit. The departure signals a structural shift in how OpenAI approaches AI safety and risk management, moving away from a dedicated safety function toward integration with broader research activities.