OpenAI is calling for the retirement of SWE-bench Verified, the widely-used AI coding benchmark, claiming most tasks are flawed enough to reject correct solutions. The company argues that leading AI models have likely seen the answers during training, meaning benchmark scores measure memorization rather than genuine coding ability.
Guide Labs has open-sourced Steerling-8B, an 8 billion parameter language model built with a new architecture specifically designed to make the model's reasoning and actions easily interpretable. The release addresses a persistent challenge in AI development: understanding how large language models arrive at their outputs.
Google is launching a free Gemini AI training program aimed at all 6 million U.S. educators. The move targets early adoption in the education sector before competitors establish similar programs.
Samsung is expanding Galaxy AI to include Perplexity as an integrated AI agent accessible via the voice command "Hey, Plex." The integration gives Perplexity access to Samsung's native apps including Notes, Clock, Gallery, Reminder, and Calendar, marking Samsung's shift toward a multi-agent ecosystem.
Cohere Labs has released tiny-aya-global, a lightweight text generation model trained to support conversational tasks across 100+ languages. The model is available on Hugging Face under a CC-BY-NC-4.0 license and builds on the tiny-aya-base architecture.
A new benchmark evaluating AI models on photo retrieval reveals significant limitations in their ability to find specific images from personal collections. The test presents models with what appears to be a simple task—locating a particular photo—yet results demonstrate the gap between general image recognition and practical personal image search.
Apple Intelligence, which automatically summarizes notifications and messages on hundreds of millions of devices, systematically generates stereotyped and hallucinated content according to an independent AI Forensics investigation. The analysis of over 10,000 AI-generated summaries reveals bias baked into the feature that pushes problematic assumptions to users unprompted.
Google has integrated a full AI Mode prompt box directly into Chrome's address bar, expanding its AI assistant capabilities in the desktop browser. This follows the company's January 2026 Gemini integration update for Chrome's side panel.
Google has restructured its AI subscription offerings into two main tiers: Google AI Pro (formerly Gemini Advanced) and the higher-tier Google AI Ultra. The tiers provide different access levels to Gemini models and capabilities, with Ultra positioned as the premium option for advanced reasoning tasks.
Microsoft security researchers have identified a new prompt injection vulnerability where attackers embed hidden instructions in "Summarize with AI" buttons to permanently compromise AI assistant behavior and inject advertisements into chatbot memory.
Anthropic is rolling out new desktop features for Claude Code that automate additional steps in the development workflow. The update extends Claude Code's capabilities beyond the web interface to provide deeper integration with local development environments.
Anthropic has released Claude Code Security, an AI tool designed to identify code vulnerabilities that traditional security scanners overlook. The announcement prompted an immediate decline in cybersecurity stock valuations.
Sarvam, an Indian AI startup, has launched Indus, a chat application currently available in beta. The move positions the company to compete against global AI players in the rapidly expanding Indian market for conversational AI.
Taalas, a new Canadian hardware startup, announced its first product: a custom silicon implementation of Meta's Llama 3.1 8B model running at 17,000 tokens/second. The startup uses aggressive quantization combining 3-bit and 6-bit parameters. The system is accessible via chatjimmy.ai.
Google's NotebookLM has received multiple updates including integration with Gemini 3.1 Pro and new prompt-based slide revision capabilities. The updates enable users to customize presentation slides through natural language prompts rather than manual editing.
GitHub has added a model picker feature to Copilot's autonomous coding agent for Business and Enterprise tier users. The feature allows teams to select which AI model powers the asynchronous background agent that handles delegated development tasks.
GitHub deprecated selected Anthropic and OpenAI models across all Copilot experiences on February 17, 2026. The deprecation affects Copilot Chat, inline edits, ask mode, agent mode, and code completions. Specific model names and transition timelines were not disclosed in the initial announcement.
Google's Gemini 3.1 Pro, an agentic coding model optimized for autonomous development workflows, is now available in public preview within GitHub Copilot. The model emphasizes efficient edit-then-test loops and tool use in early testing.
OpenAI is developing a smart speaker priced between $200 and $300 that includes a camera, facial recognition, and proactive AI recommendations. The device represents the first in a broader hardware roadmap that includes smart glasses and audio devices competing with AirPods.
OpenAI plans to release a smart speaker with an integrated camera as its first hardware device, with an estimated price between $200 and $300. The device will include object recognition, conversation detection, and a Face ID-like facial recognition system for purchases.