AI News
AI news that matters. Updated daily.
No stories match your filters.
Anthropic Refuses Pentagon Contract Over Safety Concerns
An Uncanny Valley podcast episode from February 26, 2026 covered an ongoing dispute between Anthropic and the Pentagon. According to Wired's reporting, Anthropic turned down a Department of Defense contract, with the conflict rooted in Anthropic's stated safety policies that govern what its models can be used for.
SambaNova Partners with Intel to Cut Inference Costs for Reasoning Model Workloads
SambaNova Systems announced a partnership with Intel focused on delivering AI inference infrastructure optimized for cost-effective performance on complex, multi-step reasoning workloads. The companies are targeting enterprise customers that need to run reasoning-capable models at scale where per-query cost is a significant operational factor.
Google Offers Free AI Training to Massachusetts Residents Through State Partnership
On February 26, 2026, Google and the Massachusetts AI Hub announced a partnership to provide free AI training to Massachusetts residents. The initiative is part of Google's Grow with Google program and covers foundational AI skills through online courses.
Google Translate Adds Context Buttons and Alternative Translations for Ambiguous Text
On February 26, 2026, Google announced new AI-powered features in Google Translate. The updates include alternative translation options for ambiguous text and two new interactive buttons: one labeled "understand" that provides context about a translation, and one labeled "ask" that allows users to query specific aspects of the translation.
Alphabet Reintegrates Intrinsic Robotics with Google DeepMind to Advance Physical AI
Alphabet announced that Intrinsic, its industrial robotics and automation software unit, will integrate more closely with Google DeepMind, the Gemini model family, and Google Cloud infrastructure. The reorganization is intended to accelerate the development of physical AI - AI systems that operate in and interact with the physical world in factories, warehouses, and manufacturing environments.
ServiceNow Launches AI Workforce Product That Claims to Handle Entire Job Functions
ServiceNow announced a product it is calling the Autonomous Workforce, featuring AI specialists designed to perform complete job functions rather than assist with individual tasks. The company explicitly positioned the specialists as distinct from conventional AI agents by claiming they handle entire role scopes - including judgment calls, multi-step processes, and cross-system actions - rather than just executing discrete instructions.
Google Opens Nano Banana 2 API to Developers for Image Generation and Editing
On February 26, 2026, Google announced developer access to Nano Banana 2 through its API. The model is described as Google's best image generation and editing model, combining the quality of its Pro tier with faster inference speeds.
AirSnitch Attack Bypasses Wi-Fi Encryption on Standard Home and Enterprise Networks
Security researchers disclosed a new attack called AirSnitch in late February 2026 that can bypass Wi-Fi encryption on standard home, office, and enterprise networks. The attack exploits weaknesses in how modern Wi-Fi implementations handle traffic isolation between network segments, including guest networks. Crucially, it does not require breaking encryption keys - it sidesteps them by manipulating the way access points route traffic between clients.
Wayve Raises $1.2B to Fund Commercial Autonomous Driving Trials in 2026
Wayve, a UK-based autonomous driving AI company, closed a $1.2 billion funding round. The company plans to use the capital to fund commercial trials of its autonomous driving system in 2026. Wayve's approach centers on what the company calls embodied AI - training driving models end-to-end on diverse real-world data using machine learning rather than the rules-based perception and planning modules that dominated earlier autonomous vehicle architectures.
MIT Survey: AI Spending Targets Efficiency Gains but Human-Centered Uses Deliver More Value
On February 26, MIT Technology Review Insights published research drawing on a survey of 250 industry leaders globally, examining how manufacturers and industrial companies are deploying AI under the Industry 5.0 framework. The report contrasts Industry 4.0 - which focused on integrating AI, IoT, robotics, and cloud technologies - with Industry 5.0, which the report frames as orchestrating those now-integrated technologies at scale with a specific purpose: augmenting human potential and improving sustainability outcomes rather than continuing to reduce labor costs.
OpenAI and PNNL Build Benchmark for AI-Assisted Federal Permitting Drafts
On February 26, 2026, OpenAI and Pacific Northwest National Laboratory (PNNL) announced a new benchmark called DraftNEPABench. The benchmark evaluates how well AI coding agents can assist in drafting documents for federal environmental review under the National Environmental Policy Act (NEPA).
OpenAI Codex Integrates with Figma to Connect Code and Design Workflows
On February 26, 2026, OpenAI and Figma announced a new integration connecting OpenAI Codex with the Figma design environment. The integration allows teams to generate code from designs and iterate on both the code and the canvas within a connected workflow. The stated goal is to reduce the handoff friction between designers and developers.
Hugging Face Documents MoE Improvements: 3.2x Faster Loading, Expert Parallelism
On February 26, Hugging Face published a detailed technical post on Mixture of Experts (MoE) architecture support in its transformers library. The post covers both the conceptual underpinnings of MoE models and specific engineering improvements developed to handle recent releases including DeepSeek R1, Qwen 3.5, GLM-5, and Mixtral variants.