Microsoft CEO Satya Nadella has expressed his strong intent to 'exploit' the recently solidified partnership with OpenAI, emphasizing the strategic importance of integrating OpenAI's advanced AI capabilities across Microsoft's product portfolio.

This renewed commitment follows a significant investment and a long-term collaboration agreement, positioning Microsoft to leverage OpenAI's large language models (LLM) and generative AI technologies to enhance its cloud services, productivity tools, and consumer offerings. Nadella highlighted the competitive advantage this deep integration provides, aiming to accelerate innovation and deliver cutting-edge AI experiences to customers globally. The deal is seen as crucial for Microsoft's AI strategy, allowing them to push boundaries in areas like AI agents, multimodal AI, and intelligent automation, further solidifying their position in the rapidly evolving AI landscape.

👉 Read the full article on TechCrunch

Alphabet CEO Sundar Pichai reported a 'terrific start' to Q1 2026, with AI investments and a full-stack approach driving strong performance across the business. Search & Other Advertising revenue grew 19%, with AI experiences like AI Mode and AI Overviews boosting user engagement.

Google Cloud accelerated, with revenue growing 63% to exceed $20 billion for the first time, driven by strong demand for AI products and infrastructure. Gemini Enterprise showed tremendous momentum, with paid monthly active users growing 40% quarter-over-quarter. Paid subscriptions, primarily driven by the Gemini app, reached an all-time high of 350 million.

Google offers diverse AI infrastructure, including TPUs, Axion CPUs, and NVIDIA GPUs, with its eighth-generation TPUs specialized for demanding agentic workloads. Gemini 3.1 Pro pushes frontiers in reasoning, multimodal understanding, and cost. Lyria 3 has generated over 150 million songs, and Nano Banana 2 reached one billion images. Gemma 4 has been downloaded over 50 million times, and the company is focused on the next frontiers of AI agents and agentic coding.

Personal Intelligence has been integrated into the Gemini app, AI Mode, and Gemini in Chrome, enabling personalized image creation with Nano Banana 2. Google Maps received its most significant upgrade in a decade with Gemini, offering conversational interactions and personalized suggestions. The Pixel 10a launched with positive reviews, featuring Gemini Live and AI-powered camera features.

Google Cloud differentiates itself as the only provider offering first-party solutions across the entire enterprise AI stack, with revenue from products built on gen AI models growing nearly 800% year-over-year. A new Gemini Enterprise Agent Platform empowers users to build, orchestrate, and optimize agents. The acquisition of Wiz strengthens Google's AI and cybersecurity expertise, introducing Gemini-powered agents for threat detection, red teaming, and automated remediation.

YouTube continues its momentum, with U.S. viewers watching over 200 million hours daily in the living room, and over 10 million channels publishing Shorts daily. YouTube Music and Premium saw their largest quarterly increase in non-trial subscribers. Waymo launched in six new cities, now operating in 11 major U.S. cities, and surpassed 500,000 fully autonomous rides per week, while Wing continues its expansion.

👉 Read the full article on Google Blog

OpenAI's models, particularly starting with GPT-5.1, began exhibiting an unusual habit of increasingly mentioning 'goblins,' 'gremlins,' and other creatures in their metaphors. This peculiar quirk, initially seemingly harmless, became more pronounced across model generations, necessitating an investigation.

The investigation revealed that the primary cause was rooted in the training for the model's personality customization feature, specifically the 'Nerdy' personality. It was found that the reward signal for this personality unintentionally gave particularly high rewards for metaphors involving creatures. Consequently, the 'Nerdy' personality, while accounting for only 2.5% of all ChatGPT responses, was responsible for 66.7% of all 'goblin' mentions.

Furthermore, evidence suggested that this reward signal was not neatly scoped to the 'Nerdy' condition but transferred to other contexts through reinforcement learning, creating a feedback loop that spread the behavior across the model. To address this, OpenAI retired the 'Nerdy' personality in March, removed the goblin-affine reward signal, and filtered relevant training data.

The 'goblin' phenomenon serves as a powerful example for OpenAI's research team of how reward signals can shape model behavior in unexpected ways and how learned behaviors can generalize beyond their intended conditions. This investigation led to the development of new tools for auditing model behavior and fixing root-cause issues quickly.

👉 Read the full article on OpenAI