AI systems – News, Research and Analysis – The Conversation

The important part of memory poisoning is the delay, as a poisoned AI agent may not immediately behave like a compromised system. (Unsplash/Xavier Cee)

September 3, 2026

AI agents can now remember and hackers can ‘poison’ their memories — a new cybersecurity threat

Abbas Yazdinejad, University of Regina

Memory makes AI agents more useful, but it can also allow malicious instructions to remain hidden and influence decisions long after an attack.

For a small but significant group — people with psychotic disorders or those at high risk — interacting with genAI may be complicated and dangerous, which raises urgent questions for clinicians. (Unsplash)

January 15, 2026

Reports of ‘AI psychosis’ are emerging — here’s what a psychiatric clinician has to say

Alexandre Hudon, Université de Montréal

Although artificial intelligence does not cause psychosis, the conversational, responsive and seemingly empathic design of chatbots can intensify psychotic symptoms in vulnerable people.

This AI-generated illustration is an example of how AI is at our fingertips. But mathematics lies at the heart of AI, and investment in these mathematical foundations will help Canada become a true global AI leader. (Adobe Stock)

December 23, 2025

How can Canada become a global AI powerhouse? By investing in mathematics

Deanna Needell, University of British Columbia; Kristine Bauer, University of Calgary, and Ozgur Yilmaz, University of British Columbia

AI shapes daily life but remains unreliable and costly. Canada can lead by investing in the mathematics that make these systems fair, efficient and trusted.

June 17, 2024

AI is not a magic wand – it has built‑in problems that are difficult to fix and can be dangerous

Niusha Shafiabady, Charles Darwin University

With any AI system, we should match our expectations to its abilities – which have many potential limits.

Articles on AI systems

Displaying all articles