AI news — Page 2

The four-principle paper asks organizations to match control, choice, flexibility and resilience to each workload, while offering no independent outcome evidence.

The service pairs infrastructure and IaC analysis with prioritized remediation packages, but AWS has not published independent performance evidence.

The response opens a new AI Airlock phase and sets lifecycle monitoring, draft guidance and a Spring 2027 roadmap as the next implementation checkpoints.

The proof of concept adds detectable signatures to AI-designed protein sequences and predicted structures while laboratory and model tests probe whether biological function is preserved.

The targeted review asks whether the EU framework needs additional measures and invites rights holders, AI providers, research bodies and others to respond by November 3.

The research-preview benchmark uses 219 public pull requests and a model-assisted judging pipeline, with GitHub's own audit and production experiment supplying the current validation evidence.

The result covered weekly U.S. hospital-admission forecasts, but the available official page provides no methodology table or independent performance analysis.

The September 30 package requires human review around AI-linked employment decisions, preserves clinical judgment, and extends surveillance and provenance safeguards.

Early-access customers can begin engineering work with the XQRVC1902 package now, while Class Y flight qualification remains targeted for the second half of 2027.

The beta AI Function turns unstructured text into probabilities, named choices or ordered scores for routing, classification and agent evaluation without asking a general-purpose model to generate prose.

The new open models can move between graphical interfaces, code, MCP and APIs, with weights and evaluation trajectories available for inspection.

The public beta lets Teams and Enterprise customers share a Grok Bot's tools and role knowledge while keeping each member's conversations and personal context separate.

The 8-billion-parameter model now powers a faster report mode in Asta and can run on an institution's own infrastructure, though its published evaluation is not a comparison with today's frontier models.

The ARPA-H program will test digital twins, continuous analysis and agentic operations, while three companion projects target sites, data and patient navigation.

The orbital research mission has moved from a planned hardware test to an operating satellite, while TPU radiation, thermal and spaceflight results remain ahead.

Copilot CLI and the Copilot app can operate supported desktop interfaces on macOS and Windows, with per-app approval and organization controls framing the preview.

The frontier model remains unavailable to most developers and consumers while selected Fairwind partners test a one-million-token output ceiling under restricted security controls.

The public roadmap set September 30 for two security deliverables and October 1 for a broader Constitution-alignment program, but the fetched page does not yet establish completion.

The September circular maps translation, drafting, summarization and accessibility tools that eligible public bodies can use under the European Commission's multilingual-services program.

After reviewing more than 600 comments, the U.S. standards agency selected software delivery as the first environment for testing how AI agents should be identified and authorized.

The Johnson Space Center prototype is designed to help International Space Station flight controllers navigate mission data, but no operational deployment or evaluation record is public.

The new PE12 route will evaluate foundational AI research separately from the ERC's broader computer-science panel when the next EU research framework begins.

The regulator's research finds broad exposure to AI but limited use for phone and broadband tasks, with consumers wary of losing access to people.

The Atlas platform combines agent execution with memory, retrieval and governance controls for teams building production systems on MongoDB.