May 24, 2026
- Issue 22
- 4 min read
- 14 stories / 6 sections
This issue of AI Today examines the intersection of artificial intelligence with human experience, creativity, and ethics. From Capital One Canada’s balanced approach to AI in customer service to Ottawa's efforts to combat invasive species, the diverse impacts of AI stretch across sectors and communities.
Contents (6 sections)
Canada
Canadian AI policy, companies, and adoption
- 01
reddit.com (opens in new tab)
A Toronto-based founder is seeking beta testers for ClairFlo, a free AI accounting tool.
ClairFlo automates bookkeeping, reducing the need for manual data entry.
Interested users can sign up to test the tool and provide feedback.
Policy & Regulation
Privacy, ethics, governance, regulation
- 01
Montreal author Chanel Sutherland defends her writing as human after AI detector flagged prizewinning story (opens in new tab)
Author Chanel Sutherland's story Descend was flagged by an AI detector as likely AI-generated.
The incident highlights concerns over AI bias and the ethics of policing creative writing.
Ongoing discussions in the literary community about AI's role in writing are likely to continue.

- 02
Religious groups must keep careful eye on artificial intelligence - Winnipeg Free Press (opens in new tab)
A meeting took place to discuss the role of religion in artificial intelligence development.
Religious leaders may help shape the ethical frameworks guiding AI technology.
Future discussions may influence how faith perspectives integrate into AI applications.

- 03
Volatility Is a Mandate: Why Economic Uncertainty Should Accelerate Government Reinvention (opens in new tab)
Leaders across Canada’s public sector are urged to accelerate transformation amid economic uncertainty.
Citizens demand enhanced services while governments face pressing financial and demographic challenges.
Canada’s AI strategy aims to prioritize productivity and global competitiveness for future developments.

Government & Public Sector
Federal use, public-sector AI, sovereign compute
- 01
'Use your brain': Hundreds march in Vancouver against AI data centres (opens in new tab)
Over 500 people marched in downtown Vancouver to protest the expansion of AI data centres.
The protest reflects growing community concerns about AI infrastructure's impact on local neighborhoods.
Vancouver plans data centre projects in Mount Pleasant and downtown, raising potential for further backlash.
- 02
AI Is Interested in You, So You'd Best Be Interested in It - Atalayar (opens in new tab)
AI is significantly influencing all job sectors, even traditionally secure positions like nursing and electrical work.
Many companies hesitate to hire as they fear AI will take over future job roles.
Some experts predict a sharp drop in global employment, then a potential recovery, though timelines are unclear.

Industry & Models
Investment, M&A, models, agents, coding, ASI/AGI
- 01
Exclusive Interview: "Innovate or Evaporate" — Why the AI Productivity Gap is the Biggest Threat to Canadian SMEs - The Globe and Mail (opens in new tab)
Canadian small and medium-sized enterprises are struggling to adopt artificial intelligence technologies.
This AI productivity gap threatens their competitiveness against larger firms that effectively use AI.
Experts recommend immediate action to integrate AI solutions for sustainability and growth.
- 02
Hotels strive to be found as AI models conduct travel search - France 24 (opens in new tab)
Hotels are adapting their strategies to improve visibility in AI-driven travel searches.
A significant number of travelers now use AI tools to plan trips, altering industry dynamics.
Hotels must enhance their systems to better respond to vague customer requests.

Sectors & Applications
Agriculture, environment, jobs, applied AI
- 01
Capital One Canada balancing AI use with human touch - Benefits Canada.com (opens in new tab)
Capital One Canada balances artificial intelligence and human interaction in its operations.
This approach aims to enhance customer experience while maintaining a personal touch.
- 02
Ottawa using AI to detect dangerous wild parsnip plant - Yahoo News Canada (opens in new tab)
Ottawa began using artificial intelligence to detect the dangerous wild parsnip plant.
This approach aims to protect local ecosystems and public health from invasive species.
Research
Trending AI research papers from arXiv and Hugging Face
- 01
APEX: Autonomous Policy Exploration for Self-Evolving LLM Agents (opens in new tab)
Researchers introduced APEX, a method for self-evolving large language model (LLM) agents.
APEX improves decision-making by preventing exploration collapse and enhancing sustained exploration.

- 02
[2605.21027] Beyond Text-to-SQL: An Agentic LLM System for Governed Enterprise Analytics APIs (opens in new tab)
Researchers developed Analytic Agent, a Large Language Model system for enterprise analytics APIs.
It helps non-technical users access data securely, reducing risk in complex environments.
Evaluations on real enterprise use cases show reliability in user goal interpretation and compliance.
![[2605.21027] Beyond Text-to-SQL: An Agentic LLM System for Governed Enterprise Analytics APIs](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png)
- 03
Conformal Selective Acting: Anytime-Valid Risk Control for RLVR-Trained LLMs (opens in new tab)
Researchers introduced Conformal Selective Acting, a method for real-time risk control in deployed models.
This approach ensures safety certification for each deployment round without relying on average performance metrics.

- 04
[2605.20402] Decomposing MXFP4 quantization error for LLM reinforcement learning: reducible bias, recoverable deadzone, and an irreducible floor (opens in new tab)
Researchers decompose quantization error in MXFP4 for reinforcement learning in large language models.
This analysis identifies specific error components, improving accuracy in model training and performance.
Corrections to quantization errors recover near BF16 accuracy in various model configurations.
![[2605.20402] Decomposing MXFP4 quantization error for LLM reinforcement learning: reducible bias, recoverable deadzone, and an irreducible floor](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png)