Launch editionUpdated when the evidence changes

Policy specialists reviewing an AI impact assessment

Canada · Lead story

Canada wants clearer AI disclosures. Here is what useful transparency looks like.

Labels, incident records and agent logs could change how Canadian organizations design AI products.

By AI New DeskAugust 10, 20264 min read

QUICK TAKES

Sovereign AIWhy compute is now industrial policy RAGThe work that happens before the model writes EU AI ActWhat Canadian exporters need to know

NEW: AI LEARNING LAB

Don’t just read the feed. Build an AI knowledge map.

Follow curated tracks, save a queue, test yourself with practical questions and master the terms shaping every AI story.

Enter the Learning Lab →

01Personal learning paths02Knowledge quizzes03AI flashcards04Saved reading queue

TOPIC GUIDES

Choose a path through the noise.

View every guide →

01Canadian AI policy: the practical guideA plain-language guide to Canadian AI policy, privacy, procurement, compute, safety and the institutions shaping the rules.02How to use AI well: a practical learning pathPractical AI guides for prompting, research, writing, spreadsheets, privacy, evaluation and human-reviewed agent workflows.03AI models explained: capabilities, cost and evaluationUnderstand AI model releases, reasoning, context windows, retrieval, fine-tuning, quantization, tool use and evaluation.04The plain-language AI glossaryThirty-five core terms with practical definitions and important limits.

THE HOME DESK

Canada is building its AI playbook in public.

All Canada coverage →

CanadaCanada wants clearer

CanadaCanada watch

Canada wants clearer AI disclosures. Here is what useful transparency looks like.

Labels, incident records and agent logs could change how Canadian organizations design AI products.

August 10, 20264 min read

A colourful documentary photograph of a diverse Canadian town-hall workshop mapping six AI policy pillars with cards on a wall, subtle maple-leaf context, energetic civic atmosphere, no readable text

CanadaCanada's AI for

CanadaPOLICY LENS

Canada's AI for All Strategy: A Field Guide to the Six Pillars

A close reading of Canada's 2026 national AI strategy, what its promises mean in practice, and the milestones citizens and businesses should watch next.

August 10, 20269 min read

Cinematic night photograph inside a Canadian high-performance computing centre, vivid blue and magenta server light, technicians inspecting liquid-cooled accelerator racks, recognizable but subtle northern setting, no logos

CanadaSovereign AI Compute

CanadaEXPLAINER

Sovereign AI Compute in Canada: What Capacity Actually Buys

Compute sovereignty is about much more than owning GPUs. This explainer follows the chain from electricity and chips to researchers, public-interest workloads and economic leverage.

August 9, 20268 min read

MODELS & PRODUCTS

What changed—and what’s just a launch-day claim.

Model tracker →

Playful editorial photograph of a model racetrack where benchmark scorecards are examined under a magnifying lens, colourful lab props, sophisticated visual metaphor, no numbers or readable text

ModelsHow to Read

ModelsFIELD GUIDE

How to Read an AI Benchmark Score Without Being Misled

Leaderboards compress a complicated experiment into one number. Learn to inspect the task, sample, scoring method, contamination risk and practical relevance behind it.

July 21, 20268 min read

Surreal wide-angle editorial photograph of an enormous colourful library desk feeding selected pages into a glowing AI workspace, visible foreground-to-background depth, lively and elegant, no text

ModelsContext Windows Explained:

ModelsEXPLAINER

Context Windows Explained: Capacity Is Not Memory

July 20, 20268 min read

Cinematic macro photograph of a transparent mechanical puzzle with multiple illuminated reasoning paths converging on one answer chamber, deep jewel colours, scientific editorial style, no symbols or text

ModelsReasoning Models and

ModelsEXPLAINER

Reasoning Models and Test-Time Compute, in Plain English

July 19, 20268 min read

Colour-rich candid photograph of a person using one AI assistant across voice, camera and text in a bustling kitchen workshop, visible sound and image cues as subtle light, natural human energy, no logos

ModelsMultimodal AI: What

ModelsVIDEO DEBRIEF

Multimodal AI: What Changes When a Model Can See and Hear

July 18, 20268 min read

DEEP READS

Follow the evidence.

Browse all →

$Bold editorial photograph of a business workflow built as a colourful balance scale, time savings on one side and review, errors and customer value on the other, realistic office, no text$ BusinessAI ROI: Measure

BusinessOPERATOR PLAYBOOK

AI ROI: Measure the Whole Workflow or Fool Yourself

Time saved in one step is not business value. A credible case includes review, integration, errors, adoption, delay and the outcome customers actually experience.

June 11, 20268 min read

Colourful editorial photograph of a four-week experiment board with real workflow stations, mixed team testing and a clear decision gate at the finish, energetic documentary style, no text

BusinessDesign a Four-Week

BusinessOPERATOR PLAYBOOK

Design a Four-Week AI Pilot That Can Produce a Real Decision

A strong pilot has a baseline, representative tasks, stop conditions and a named decision at the end - not a collection of enthusiastic anecdotes.

June 10, 20268 min read

Stylish editorial photograph of three AI vendor machines opened for inspection on equal test benches, evaluators checking evidence, locks, cost and exit ports, vivid colours, no logos

BusinessThe AI Vendor

BusinessFIELD GUIDE

The AI Vendor Scorecard That Goes Beyond the Demo

Compare evidence, data terms, security, evaluation access, accessibility, price and exit - using your own tasks before a contract narrows your options.

June 9, 20268 min read

Dramatic editorial still-life of an AI service contract as ten colourful interlocking protective layers around data and a clear exit door, premium legal desk photography, no readable text

BusinessTen AI Contract

BusinessFIELD GUIDE

Ten AI Contract Clauses That Matter After the Sales Demo

Data use, subprocessors, model changes, evaluation, incidents, indemnity and exit terms determine whether a promising pilot remains governable.

June 8, 20268 min read

HOW WE REPORT

Primary sources first. Clear labels. Corrections in public.

AI moves fast; accuracy still matters. Our news analysis links to the original announcement and separates reported fact, company claim and AI New analysis.

Read our standards →