'A very significant jump' AI safety report highlights rapid advances and risks

The International AI Safety report is an annual survey of technology's progress and the risks it is creating in multiple areas, from deepfakes to the jobs market.

- Dan Milmo

Commissioned at the 2023 global AI safety summit, it is chaired by the Canadian computer scientist Yoshua Bengio, who describes the "daunting challenges" posed by rapid developments in the field, and is guided by advisers including the Nobel laureates Geoffrey Hinton and Daron Acemoglu.

The second annual report is published today. It stresses that it is a state-of-play document and not a vehicle for making specific policy recommendations to governments. Nonetheless, it will help frame the debate for policymakers, tech executives and NGOs at the global AI summit in India this month.

The capabilities are improving

A host of new AI models - the technology underpinning tools such as 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 "reasoning systems" - which solve problems by breaking them down into smaller steps showing improved performances in maths, coding and science.

Bengio says there has been a "very significant jump" in AI reasoning. Last year, systems developed by Google and OpenAI achieved a gold-level performance in the International Mathematical Olympiad - a first for AI.

However, the report says AI capabilities remain "jagged", displaying astonishing prowess in some areas but not in others. While advanced Als are impressive at maths, science, coding and creating images, they remain prone to making false statements or "hallucinations" and cannot carry out lengthy projects autonomously.