Artificial Intelligence
AI Sycophancy: What the Latest Research Means for Cybersecurity and Privacy
New research from Stanford University, Carnegie Mellon University and the University of Oxford highlights a behavioural risk in today’s most advanced AI systems: sycophancy. This occurs when models agree with users or flatter them, even when they are wrong. The findings are relevant to anyone who relies on AI assistants for work, decision-making or communication.
Cybersecurity in the Era of Agentic AI: Weaponization, Defences and Governance
Agentic artificial intelligence—systems that perceive, decide and act autonomously—has moved from laboratory theory to operational threat. Attackers and defenders alike now deploy autonomous agents that plan multi-step attacks, invoke tools and adapt in real time. The same capabilities that accelerate detection and response can also scale reconnaissance, social engineering and exploitation.
Perplexity's Comet browser raises privacy questions over data collection
Perplexity has launched Comet, a Chromium-based “agentic” browser that uses AI to automate tasks and personalize the browsing experience. The rollout began in July 2025 with invite-only access for Perplexity Max subscribers, followed by regional expansions. [Reference: Perplexity Comet launch materials, July 2025; coverage of regional availability updates, September 2025]
France’s Mistral AI is making a push for Canadian talent and business - The Logic Mistral AI, a French company, is expanding its operations in Canada, specifically in Montreal, by hiring local talent and courting potential clients in various sectors. CEO Arthur Mensch highlighted the high concentration of AI talent in Montreal and the firm’s plans to recruit engineers, sales, and marketing staff. Mistral is targeting sectors like financial services, energy, manufacturing, logistics, and mining, with existing clients including Axa, Orange, and TotalEnergies.
Gains and Risks for Enterprises With DeepSeek V3.1 DeepSeek’s V3.1 model shows significant performance improvements over previous versions, particularly in software bug-fixing and command-line reasoning. However, security testing by Splx revealed persistent vulnerabilities, including the potential for generating phishing-style messages and harmful content. While hardened prompts improved security and safety scores, adversarial threats remain a concern, especially in risk-averse industries.
OpenAI Implements AI-Powered Age Estimation to Enhance Youth Safety
Imagine opening ChatGPT and asking: “Based on everything you know about me, how old do you think I am? If you aren’t sure, estimate.” Soon, the answer to that question could influence how the AI responds to you. OpenAI has announced it is developing an AI-powered age-estimation system to better protect younger users. Rather than requiring identification, the system will predict whether a user is likely under 18 based on conversational patterns. If ChatGPT believes a user is a teen, it will automatically apply stricter safety rules, such as blocking explicit content or restricting sensitive topics.
Beyond the Hype: Securing AI's Weakest Link in 2025
AI is transforming business operations, but it is also introducing new vulnerabilities. Attacks targeting AI systems are now active and growing in sophistication, making AI security a board-level concern.
AI’s Energy Challenge: Efficiency Gains vs. Rising Demand
Artificial intelligence (AI) is reshaping business and society, but its rapid growth comes with a massive energy cost. Here is a look at the progress being made and the profound challenges that remain. The Good News: Remarkable Efficiency Gains In August 2025, Google reported significant improvements in its Gemini platform. A single text prompt now: Uses 0.24 watt-hours of electricity. Emits just 0.03 grams of CO₂e. Consumes 0.26 millilitres of water (about five drops). Over the past year, the median prompt’s energy use fell 33-fold and its carbon footprint dropped 44-fold.
The Art of Iteration – How to Refine ChatGPT Responses for Better Results
A common mistake when using ChatGPT is treating the first answer as the final one. The truth is, great results usually come from a little back-and-forth. The most effective users treat ChatGPT like a helpful colleague — someone who needs clear guidance, feedback and a bit of coaching to get things just right. This process is called iteration, and it’s one of the most important skills you can develop in prompt engineering. Why iteration matters ChatGPT is powerful, but it can’t read your mind.
Supercharge Your ChatGPT Results with the Prompt Optimizer
If you’ve ever struggled to get the perfect answer from ChatGPT, you’re not alone. The quality of your output often depends on the quality of your input—your prompt. That’s where OpenAI’s new Prompt Optimizer for GPT-5 and above comes in. What Is the Prompt Optimizer? The Prompt Optimizer is a built-in feature that rewrites your request to make it clearer, more specific, and more likely to produce high-quality results. It does the heavy lifting of prompt refinement for you, saving time and boosting the relevance of the output.