“Deepfakes are spreading and more AI companions”: seven insights from the latest security report on artificial intelligence

1. The capabilities of AI models are improving

A variety of new AI models — the technology that underlies tools like chatbots — were released last year, including OpenAI’s GPT-5, Anthropic’s Claude Opus 4.5 and Google’s Gemini 3. The report points to new “inference systems” – which solve problems by breaking them down into smaller steps – showing improved performance in math, coding and science. Bengio said there has been a “very significant leap” in AI thinking. Last year, systems developed by Google and OpenAI achieved a gold performance at the International Mathematical Olympiad – a first for AI.

However, the report says AI capabilities remain “jagged,” referring to systems that show amazing capabilities in some areas but not others. While advanced AI systems are impressive in math, science, coding and image generation, they are still prone to making false statements or “hallucinations” and cannot complete lengthy projects autonomously.