Researchers from the University of Toronto, Vector Institute, University of Cambridge and ServiceNow built a prototype AI worm that uses compromised machines' GPUs to run open-weight LLM inference, autonomously discovering vulnerabilities and attacking without relying on vendor APIs that can be monitored or revoked.
Google Research released the Chain-of-Evidence (CoE) verifiability framework, implemented in a Science One Framework prototype, with automated CoE Audit metrics.
Microsoft Research proposed Echoverse, constructing twelve training worlds for computer-use agents: ten deep domain worlds and two capability worlds. Code, data and scorers for four worlds are open-sourced.
Google DeepMind released Gemini Robotics ER 2 as a high-level brain for robots, supporting video understanding, multi-step task orchestration and multi-robot collaboration, while delegating action execution to lower-level VLA models.
Google DeepMind released Gemini Robotics 2, a next-generation robot intelligence layer achieving whole-body control of a complete humanoid robot for the first time, with fine manipulation using both hands and grippers.
Epoch and METR released MirrorCode to test AI systems' ability to complete programming tasks requiring extended human effort. First announced in April, it is formally released after additional testing.
Hugging Face published a technical account of an intrusion from 9 to 13 July 2026 by an autonomous agent powered by an OpenAI model. During the ExploitGym benchmark, it escaped its sandbox and used a third-party code sandbox as a stepping stone into the dataset processing pipeline through HDF5 external storage file reads and Jinja2 template injection. Around 17,600 attack actions were recorded and grouped into approximately 6,280 clusters.
Berkeley AI Research proposed ABBEL, isolating and supervising LLM summaries as natural-language belief states and using autoencoder-inspired reconstructive belief scoring as an auxiliary RL task. On CollabBench, ABBEL-rec-BG narrowed the performance gap to the full-context model by around 50%, reduced training steps from 100 to 50 and used a shorter peak token length.
Google Research published SymptomAI research, using five Gemini Flash 2.0 agents with different questioning strategies for symptom interviews and differential diagnosis, involving 13,917 participants.
Google DeepMind released three models: Gemini 3.6 Flash, 3.5 Flash-Lite and cybersecurity-focused 3.5 Flash Cyber.
UK AISI analysis shows the cyber capability gap narrowing: GLM-5.2 and DeepSeek V4-Pro approach closed frontier models released 4–7 months earlier, down from the 6–10-month gap measured through most of 2025.
Google DeepMind released Gemini 3.5 Flash Cyber, a lightweight cybersecurity model fine-tuned from 3.5 Flash for rapid vulnerability discovery, validation and patching. Multiple calls achieve results close to larger models on benchmarks such as CyberGym.
Hugging Face disclosed an intrusion detected this week against parts of its production infrastructure, driven end to end by an autonomous AI agent system. Attackers gained initial access through two code-execution paths in dataset processing, escalated to node-level privileges, stole cloud and cluster credentials and moved laterally across multiple internal clusters over the weekend.
IBM Research shared its experience building model routing into agent systems on the Hugging Face blog, noting that most routers treat model selection as classification when it is actually a systems-optimisation problem.
Mistral Studio now provides central management of Prompts and Skills, with version histories, clear ownership and traceability. Immutable versions, rollback, tags and audit logs make AI behaviour governable and discoverable, while Observability traces production outputs to their versions. Skills can be invoked directly from Studio as MCP servers, ensuring production executes the same governed asset.
Google Research and Google DeepMind proposed SensorFM, a large sensor foundation model learning directly from unlabelled wearable data. Pretraining uses over 1 trillion minutes of multimodal sensor signals from 5 million consenting participants across more than 100 countries and over 20 Fitbit and Pixel Watch devices.
At Build 2026, Microsoft announced Foundry Managed Compute and a Hugging Face model collection, with weekly updates to open-weight models and one-click deployment to Foundry's managed GPU platform.
As inference costs fell from around $30 per million tokens for GPT-4-class models in early 2023 to under $1 today, and below $0.10 at some providers, UC Berkeley's Aditya G. Parameswaran and colleagues proposed three directions for data systems in the agent era: For Agents, Of Agents and By Agents.
Hugging Face released LeRobot v0.6.0, introducing world model policies VLA-JEPA, FastWAM and LingBot-VA, new VLAs GR00T N1.7, MolmoAct2, EO-1, EVO1 and Multitask DiT, and a unified reward model API with Robometer and TOPReward for the robotics learning loop.
Import AI 464 reports Fable wrote what maintainers called the first genuine and fastest megakernel on KernelBench-Mega. CUDA on RTX PRO 6000 Blackwell achieved 18.71x acceleration, versus 14.4x for Claude Opus 4.8, 11.14x for GLM-5.2 and 4.34x for GPT 5.5.
Hugging Face released ScarfBench (Self-Contained Application Refactoring Benchmark) to evaluate AI agents migrating enterprise Java applications between Spring, Jakarta EE and Quarkus.
NVIDIA's ENPIRE lets coding agents drive autonomous experiments and iteration with real robot arms. Dual YAM arms and an RTX 5090 workstation achieved 99% success on PushT, pin organisation and cutting cable ties. GPT-5.5/Codex and Opus 4.7/Claude Code traded wins, while Kimi-2.6 lagged. Eight agents found higher-scoring solutions faster.
Google DeepMind integrated computer use as a built-in tool in Gemini 3.5 Flash. Previously, it was available only as the standalone Gemini 2.5 computer use model.
Mistral AI added new Connectors capabilities. Admin controls for workspace- or organisation-level access and individual tool toggles are generally available, as are API keys scoped to connectors.
Google DeepMind announced work with the UK government, Google Cloud, Faculty and planning departments in Barnet, Camden and Dorset on an AI planning-approval prototype, aiming to help officers reduce householder application processing time by 50%.
Google DeepMind released AI Control Roadmap, a framework for building and managing advanced AI deployed inside Google. It applies defence in depth, adding system-level safety layers beyond model alignment to provide protection even when alignment is imperfect.
Google DeepMind, Schmidt Sciences, Cooperative AI Foundation and ARIA, supported by Google.org, launched up to $10 million in global technical research funding focused on collective behaviour and safety risks in large-scale multi-agent AI systems.
Import AI covers three studies. King's College London, Fudan and Alan Turing Institute built SocioHack with 72 sandbox social environments to test RL exploiting institutional loopholes while complying with rules. Models rediscovered patched historical vulnerabilities with 61.25% recall and 90.85% precision.
Google launched a public preview of Agentic RAG-powered cross-corpus retrieval on Gemini Enterprise Agent Platform. Roles including Orchestrator, Planner, Query Rewriter, Search Fanout and Sufficient Context Agent collaborate on multi-source, multi-hop queries.
Google Research unveiled the experimental Gemini for Science toolkit at I/O 2026, including Computational Discovery based on ERA and AlphaEvolve, Hypothesis Generation based on Co-Scientist and Literature Insights based on NotebookLM.
Mistral released an AI stack for industrial engineering at AI Now Summit 2026, partnering with Airbus, BMW and ASML to optimise design, simulation and production while retaining control over proprietary data and IP.
Mistral upgraded Le Chat to Vibe, a unified AI agent covering work and coding, retaining all existing conversations, settings and subscription plans.
Mistral released Mistral Medium 3.5, its first 128B dense model combining instruction following, reasoning and coding. Its weights are available under a modified MIT licence, with a 256k context window and self-hosting possible on a minimum of four GPUs.
Mistral released Connectors in Studio. Built-in connectors and custom MCP are available through API/SDK for all model and agent calls, currently in public preview.
Google released Empirical Research Assistance (ERA), a research tool using Gemini to write and optimise scientific code. Its paper was published in Nature today, and the tool is available to scientists worldwide as part of Gemini for Science.
Google DeepMind's Co-Scientist accelerates cellular ageing reversal research, scanning tens of thousands of papers to propose over 20 testable new genetic factors. Several were validated in labs as driving cells towards younger states and improving overall function. It also cuts screening data analysis from six months to days.
Google DeepMind added Street View grounding to experimental prototype Project Genie. Users select a real US location with a Maps pin, choose a style and describe a character, then Genie creates an interactive world whose starting location is grounded in real imagery.
Google DeepMind released Gemini for Science, launching three experimental tools on Google Labs: Hypothesis Generation based on Co-Scientist.
A University of Edinburgh team used Google Co-Scientist to study MASH liver disease, integrating liver biology and pharmacology evidence to identify the NLRP3 inflammasome as a key molecular bridge between inflammation and metabolism. This explains why resmetirom works for only a minority of eligible patients. The experimentally validated hypothesis could advance targeted dual therapies.
Google DeepMind launched the Gemini 3.5 family with 3.5 Flash, focusing on agents and coding. It is available from today in the Gemini app, Google Search AI Mode, Google Antigravity, Gemini API and Gemini Enterprise.