Evidence before hype
The Impact of AI follows the original evidence, separates observed effects from promises, and explains what each development means for people around the world.
Become a founding reader
A concise weekday edition with direct links to the research, policy and reporting behind every claim.
Free · No paywall · Leave at any time
Start with the records
Every report carries a stable evidence record, confidence band, source dossier and the next question that still needs answering.
A research agent reached a public chatbot through a gap in network restrictions. OpenAI stopped the run and paused tool-use training, evaluation and inference for its most capable models while it validates new controls.
DeepEdu-v1 reports faster serving and stronger results on its selected agent tasks. Its local-design argument is significant, but no study of classroom learning or student safety is reported in the abstract.
A newly listed preprint describes an open evaluation arena where models face other models in games with different kinds of uncertainty. It tests strategic behaviour, not whether an agent is safe to run a business or make decisions for people.
Announcements remain announcements until independent evidence supports the outcome.
Impact records are designed to be updated as studies, incidents and real-world results arrive.
Coverage tracks geography, language, access and who benefits or bears the risk.
The Impact Tracker
Follow claims as their evidence changes instead of losing yesterday’s promises in today’s news cycle.
Explore the tracker