
AI safety after the resignations: what should we actually test?
A departure can focus attention. The harder work is deciding which risks are measurable and who gets to inspect the results.
Read the full story →Up to 50 links are picked up from the sources as they publish, with the original date and publisher shown. Feed links lead to the source; this list is not represented as AI Speed reporting.
In-depth guideAI in Canada: strategy, research and the infrastructure behind itA sourced guide to the organizations, compute systems and policy questions shaping Canada's AI sector.Read the guide →Each story separates reported developments from our analysis, links to original sources and explains what Canadian readers can check next.

A departure can focus attention. The harder work is deciding which risks are measurable and who gets to inspect the results.
Read the full story →
The benchmark contest is real. The decision facing a business or developer is more specific than which model appears first on a chart.
Read the full story →
The question is no longer whether AI uses power. It is whether grids, projects and customer demand can keep pace with one another.
Read the full story →
A promising molecule is only the start of a long, carefully tested path to patient care.
Read the full story →
A useful agent can carry out a task. A responsible team still defines “done” and checks what happened.
Read the full story →A broadband announcement can describe a funding decision, a target or a network that is already available. Those are different things. Before comparing providers, check the project’s location, the date and whether the source says service is live at your own address. The National Broadband Map is a starting point, followed by the provider’s address checker. A CRTC 50/10 target describes a policy objective; it does not certify an individual home or promise that a browser test will show those rates.
An AI research feed entry is different from a product launch. arXiv lists research submissions; a preprint may change after review and a paper’s benchmark is not proof that its model works for your business. Look for the method, limitations and later published version. Our Canadian AI guide explains how to ask about computing capacity, data handling and oversight before buying a tool. Network performance matters for remote services, but a result on our speed test measures only your route to AI Speed.
Items below are dated source records with links to their publishers, not independently reported AI Speed articles. We keep each excerpt short and label the source so you can check the original claim.

arXiv:2609.25010v1 Announce Type: new Abstract: Marketers increasingly use large language models (LLMs) as "synthetic personas" to predict how an audience will react to a piece of copy before it ships, encouraged by evidence that
Source: arXiv AI research · Original ↗
arXiv:2609.25013v1 Announce Type: new Abstract: Tabular foundation models have recently shown strong potential for structured biomedical data analysis. Among them, TabPFN has emerged as an effective approach for low-data tabular c
Source: arXiv AI research · Original ↗
arXiv:2609.25036v1 Announce Type: new Abstract: Dynamic Gaussian Splatting provides an explicit representation of evolving 3D scenes, but existing approaches are primarily optimized for reconstruction, future-state generation, or
Source: arXiv AI research · Original ↗
arXiv:2609.25088v1 Announce Type: new Abstract: Survival prediction for glioblastoma multiforme (GBM) demands models that are both accurate and interpretable, yet existing approaches treat these objectives as com- peting, where pe
Source: arXiv AI research · Original ↗
arXiv:2609.25165v1 Announce Type: new Abstract: In this report, we introduce \textbf{Ovis-Embedding}, a state-of-the-art omni-modal embedding family built on native integration of text, image, video, and audio. Instead of assembli
Source: arXiv AI research · Original ↗
arXiv:2609.25187v1 Announce Type: new Abstract: Task planning bridges high-level instructions and executable behavior in long-horizon manipulation, yet modern Vision-Language-Action (VLA) systems often leave this intermediate stru
Source: arXiv AI research · Original ↗
arXiv:2609.25199v1 Announce Type: new Abstract: Lean Pool is a repository of formalized mathematics. It is grown, maintained and optimized by AI agents.
Source: arXiv AI research · Original ↗
arXiv:2609.25254v1 Announce Type: new Abstract: Analyzing neuroimaging data requires specialized coding and statistical expertise, which limits accessibility for researchers without computational backgrounds. We present the AI Neu
Source: arXiv AI research · Original ↗
arXiv:2609.25272v1 Announce Type: new Abstract: Prospective stroke risk stratification in primary care is challenging because early risk signals are distributed across routine biomarkers and unstructured clinical narratives. We pr
Source: arXiv AI research · Original ↗
arXiv:2609.25284v1 Announce Type: new Abstract: A social agent's most basic decisions (should I react to this post? who should I reach out to?) are not purely content problems. The right action often hinges on the latent relations
Source: arXiv AI research · Original ↗
arXiv:2609.25285v1 Announce Type: new Abstract: Finding circuits in language models usually means running many careful interventions. We try something simpler: treat attention as a routing map from one forward pass, keep a small s
Source: arXiv AI research · Original ↗
arXiv:2609.25286v1 Announce Type: new Abstract: Structured-data agents in enterprise settings must reason over complex data environments whose relevant evidence is distributed across schemas, relationships, policies, and recurring
Source: arXiv AI research · Original ↗
arXiv:2609.25299v1 Announce Type: new Abstract: On repeated work, agents are inconsistent. We ran 42 tasks three times each and found that, depending on the model, 38% to 74% returned answers that did not agree. Consistency is wha
Source: arXiv AI research · Original ↗
arXiv:2609.25303v1 Announce Type: new Abstract: Large language models are increasingly deployed as course-specific tutors, but their usefulness depends on grounding in vetted instructional materials that are often revised mid-seme
Source: arXiv AI research · Original ↗
arXiv:2609.25337v1 Announce Type: new Abstract: Dialogue failures in language models are usually framed as memory failures: context too long, summaries lossy, a constraint forgotten. We argue this misses a deeper problem: in many
Source: arXiv AI research · Original ↗
arXiv:2609.25366v1 Announce Type: new Abstract: Chain-of-thought (CoT) monitoring is only meaningful if written reasoning causally constrains the answer. We introduce continuation-based causal testing, an ablation-patch interventi
Source: arXiv AI research · Original ↗
arXiv:2609.25400v1 Announce Type: new Abstract: Failure-based textual knowledge distillation aims to discover gaps in a model's knowledge by examining its task errors. The distilled knowledge can be useful for the reasoning of bot
Source: arXiv AI research · Original ↗
arXiv:2609.25405v1 Announce Type: new Abstract: Modern information retrieval increasingly employs both embedding and generative models to handle complex queries. However, current serving systems suffer from low throughput and poor
Source: arXiv AI research · Original ↗
arXiv:2609.25408v1 Announce Type: new Abstract: Online A/B experiments are the decision standard for user engagement, but traffic and readout time limit how many conversational-AI changes can be tested. We ask whether an offline s
Source: arXiv AI research · Original ↗
arXiv:2609.25443v1 Announce Type: new Abstract: Moving authorization earlier can shorten an agent's dispatch boundary without removing authorization work. It can also admit an action whose payload, authority, or relevant state has
Source: arXiv AI research · Original ↗
Growing industry support demonstrates a shared commitment to responsible data centre development in Canada
Source: Innovation, Science and Economic Development Canada · Original ↗
Government of Canada announces major contribution to strengthen Canada’s dual-use life sciences sector
Source: Innovation, Science and Economic Development Canada · Original ↗
The Honourable Evan Solomon, Minister of Artificial Intelligence and Digital Innovation and Minister responsible for the Federal Economic Development Agency for Southern Ontario, and Karsten Wildberger, Germany’s Minister for Digi
Source: Innovation, Science and Economic Development Canada · Original ↗
Canada’s flagship AI conference highlighted new investments, international partnerships and the country’s leadership in AI
Source: Innovation, Science and Economic Development Canada · Original ↗
Funding will support the development of a new safe-by-design AI
Source: Innovation, Science and Economic Development Canada · Original ↗
Minister Solomon to participate at ALL IN 2026
Source: Innovation, Science and Economic Development Canada · Original ↗
Canada and Alberta are expanding high-speed Internet access in the province
Source: Innovation, Science and Economic Development Canada · Original ↗
Investment from federal and provincial governments will bring high-speed Internet access to 1,833 homes across Alberta
Source: Innovation, Science and Economic Development Canada · Original ↗
Government of Canada welcomes major new investment in sovereign AI infrastructure in Saskatchewan
Source: Innovation, Science and Economic Development Canada · Original ↗
New national partnership with Amii will provide free, practical AI learning for students, educators, workers and communities across Canada
Source: Innovation, Science and Economic Development Canada · Original ↗
Extension will continue to save Canadians 10 cents per litre on gasoline and 4 cents per litre on diesel
Source: Innovation, Science and Economic Development Canada · Original ↗
Government of Canada to launch Canada’s National AI Literacy Initiative
Source: Innovation, Science and Economic Development Canada · Original ↗
Canada advances international collaboration at the G20 Innovation Ministerial
Source: Innovation, Science and Economic Development Canada · Original ↗
New framework commits data centres to protecting electricity ratepayers, limiting water use and delivering lasting benefits for Canadians
Source: Innovation, Science and Economic Development Canada · Original ↗
Small businesses are the backbone of Canada’s economy, but many are facing the pressure of United States (U.S.) tariffs on the workers they employ, the communities they support and the investments they make.
Source: Innovation, Science and Economic Development Canada · Original ↗
The Honourable Evan Solomon, Minister of Artificial Intelligence and Digital Innovation and Minister responsible for the Federal Economic Development Agency for Southern Ontario, will make an announcement on the responsible develo
Source: Innovation, Science and Economic Development Canada · Original ↗
The Honourable Rechie Valdez, Minister of Women and Gender Equality and Secretary of State (Small Business and Tourism), will highlight new measures to protect and support Canadian workers and businesses.
Source: Innovation, Science and Economic Development Canada · Original ↗
Government of Canada invests in Xanadu to build up advanced quantum manufacturing in Canada
Source: Innovation, Science and Economic Development Canada · Original ↗
Parliamentary Secretaries Bardeesy and Chi to highlight major investments to attract top researchers to Canada
Source: Innovation, Science and Economic Development Canada · Original ↗
Minister Joly announces over $99 million investment in Canada’s life sciences sector
Source: Innovation, Science and Economic Development Canada · Original ↗Canadian announcements come from ISED and the CRTC. Global AI research comes from arXiv and is labeled separately; a preprint has not necessarily been peer reviewed. Feed summaries remain short and link back to the original publisher. Longer AI Speed articles are written as distinct guides with references, rather than generated from each incoming headline.