OpenAI unveiled GPT-Rosalind, a specialized frontier model designed to accelerate drug discovery and genomics research, named after chemist Rosalind Franklin. The model is already being tested with Amgen, Moderna, and Thermo Fisher, and scored above the 95th percentile of human experts on RNA sequence-to-function prediction tasks in partnership with Dyno Therapeutics. Access is gated through a trusted-access program for qualified enterprise customers.
The White House is preparing to give major federal agencies access to Anthropic's Mythos model — the same system the Pentagon blacklisted as a "supply chain risk" over disputes about AI safeguards. Mythos has reportedly found thousands of major software vulnerabilities, making it both a powerful cybersecurity tool and a potential risk. Anthropic CEO Dario Amodei is scheduled to meet White House chief of staff Susie Wiles on Friday to try to resolve the standoff.
OpenAI has agreed to spend over $20 billion on Cerebras servers over three years and will receive an equity stake via warrants for up to 10% of the company. Cerebras is expected to refile IPO paperwork imminently, targeting a $35 billion valuation — a 60% jump from its $22 billion February valuation. The deal signals OpenAI's aggressive push to diversify away from Nvidia dependency.
Google released Gemma 4, a family of open-weight models built from the same research as Gemini 3, under an Apache 2.0 license. The 31B dense model claimed third place on Arena AI's text leaderboard, beating models 20 times its size. The family spans four variants from 2B to 31B parameters with 256K context, native vision/audio, and support for 140+ languages — targeting everything from Raspberry Pi to cloud data centers.
A new PwC study of 1,217 senior executives across 25 sectors finds that a small elite of companies generates 7.2x more AI-driven gains than average competitors. The key differentiator is not volume of AI deployment but strategy: leaders use AI for growth and new revenue streams, not just cost-cutting. AI leaders are increasing autonomous decision-making at nearly 3x the rate of peers.
The Stanford AI Index Report 2026 finds that the best AI agents perform only about half as well as PhD-level human experts on complex scientific workflows. While researchers are embracing AI tools — 6-9% of natural science publications now mention AI — there's a noted paradox: scientists using AI produce more papers but on a narrower set of topics.
The three leading US AI labs are sharing information through the Frontier Model Forum to clamp down on Chinese competitors extracting results from their cutting-edge models. This marks an unusual instance of direct collaboration among fierce rivals, driven by shared concerns over intellectual property protection and national security implications.
