However, existing frameworks face several interrelated limitations.

Research has suggested super-recognisers look at more areas across a face than typical people. Research has suggested super-recognisers look at more areas across a face than typical people. Research uses eye-tracking data to examine some people's extraordinary recognition ability They have been used in the search for the Salisbury novichok poisoners, finding murder suspects and even spotting sexual predators. Now, research has revealed fresh insights into why super-recognisers are so good at identifying faces. Previous research has suggested people with an extraordinary ability to recognise people look at more areas across a face than typical people.

Consider a bandit algorithm that recommends actions to self-interested users in a recommendation system. The users are free to choose other actions and need to be incentivized to follow the algorithm's recommendations.

Traditionally, these would be made according to hard-coded expert heuristics implemented in solvers.

Reinforcement learning (RL) is increasingly applied to real-world problems involving complex and structured decisions, such as routing, scheduling, and assortment planning. These settings challenge standard RL algorithms, which struggle to scale, generalize, and exploit structure in the presence of combinatorial action spaces. We propose Structured Reinforcement Learning (SRL), a novel actor-critic framework that embeds combinatorial optimization layers into the actor neural network. We enable end-to-end learning of the actor via Fenchel-Young losses and provide a geometric interpretation of SRL as a primal-dual algorithm in the dual of the moment polytope. Across six environments with exogenous and endogenous uncertainty, SRL matches or surpasses the performance of unstructured RL and imitation learning on static tasks and improves over these baselines by up to 92% on dynamic problems, with improved stability and convergence speed.

Anthropic announced that its Claude AI can integrate with Google Workspace. This tie-in allows the AI assistant to access any information in Gmail, Google Documents and Google Calendar. Enterprise-level customers even get a special cataloguing option for Documents that aims to offer even better speed and accuracy when retrieving information. This update could make Claude more helpful when it comes to using the chatbot for scheduling or accessing information within the Google ecosystem. The blog post with the announcement specified that the Enterprise option comes with special security controls for confidentiality, but doesn't detail if or how other users might be able to keep Claude from accessing sensitive information that might be stored in an email or document.

The modern enterprise is facing an unprecedented surge in digital identities, with machine identities now significantly outnumbering human identities. This paper examines the cybersecurity risks emerging from what we define as the "human-machine identity blur" - the point at which human and machine identities intersect, delegate authority, and create new attack surfaces. Drawing from industry data, expert insights, and real-world incident analysis, we identify key governance gaps in current identity management models that treat human and machine entities as separate domains. To address these challenges, we propose a Unified Identity Governance Framework based on four core principles: treating identity as a continuum rather than a binary distinction, applying consistent risk evaluation across all identity types, implementing continuous verification guided by zero trust principles, and maintaining governance throughout the entire identity lifecycle. Our research shows that organizations adopting this unified approach experience a 47 percent reduction in identity-related security incidents and a 62 percent improvement in incident response time. We conclude by offering a practical implementation roadmap and outlining future research directions as AI-driven systems become increasingly autonomous.

Isla Fulford, a researcher at OpenAI, had a hunch that Deep Research would be a hit even before it was released. Fulford had helped build the artificial intelligence agent, which autonomously explores the web, deciding for itself what links to click, what to read, and what to collate into an in-depth report. OpenAI first made Deep Research available internally; whenever it went down, Fulford says, she was inundated with queries from colleagues eager to have it back. "The number of people who were DMing me made us pretty excited," says Fulford. Since going live to the public on February 2, Deep Research has proven to be a hit with many users outside the company too.